docs(plugins): cover every pluggable surface in both the overview and how-to

Both plugins.md and build-a-hermes-plugin.md now cover every extension
surface end-to-end \u2014 general plugin APIs, specialized plugin types,
config-driven surfaces \u2014 with concrete authoring patterns for each.

plugins.md:
- 'What plugins can do' table grows from 9 rows (general ctx.register_*
  only) to 14 rows covering register_platform, register_image_gen_provider,
  register_context_engine, MemoryProvider subclass, register_provider
  (model). Each row links to its full authoring guide.
- New 'Plugin sub-categories' section under Plugin Discovery explains
  how plugins/platforms/, plugins/image_gen/, plugins/memory/,
  plugins/context_engine/, plugins/model-providers/ are routed to
  different loaders \u2014 PluginManager vs the per-category own-loader
  systems.
- Explicit mention of user-override semantics at
  ~/.hermes/plugins/model-providers/ and ~/.hermes/plugins/memory/.

build-a-hermes-plugin.md:
- New '## Specialized plugin types' section (5 sub-sections):
  - Model provider plugins \u2014 ProviderProfile + plugin.yaml example,
    auto-wiring summary, link to full guide
  - Platform plugins \u2014 BasePlatformAdapter + register_platform() skeleton
  - Memory provider plugins \u2014 MemoryProvider subclass example
  - Context engine plugins \u2014 ContextEngine subclass example
  - Image-generation backends \u2014 ImageGenProvider + kind: backend example
- New '## Non-Python extension surfaces' section (5 sub-sections):
  - MCP servers \u2014 config.yaml mcp_servers.<name> example
  - Gateway event hooks \u2014 HOOK.yaml + handler.py example
  - Shell hooks \u2014 hooks: block in config.yaml example
  - Skill sources (taps) \u2014 hermes skills tap add example
  - TTS / STT command templates \u2014 tts.providers.<name> with type: command
- Distribute via pip / NixOS promoted from ### to ## (they were orphaned
  after the reorganization)

Each specialized / non-Python section has a concrete, copy-pasteable
example plus a 'Full guide:' link to the authoritative doc. Devs arriving
at the build-a-hermes-plugin guide now see every extension surface at
their disposal, not just the general tool/hook/slash-command surface.

Verified:
- Docusaurus build SUCCESS, zero new broken links
- All new cross-links (developer-guide/model-provider-plugin,
  adding-platform-adapters, memory-provider-plugin, context-engine-plugin,
  user-guide/features/mcp, skills#skills-hub, hooks#gateway-event-hooks,
  hooks#shell-hooks, tts#custom-command-providers,
  tts#voice-message-transcription-stt) resolve
- Same 3 pre-existing broken links on main (cron-script-only, llms.txt,
  adding-platform-adapters#step-by-step-checklist)

AGENTS.md

@@ -42,6 +42,7 @@ hermes-agent/
 ├── plugins/              # Plugin system (see "Plugins" section below)
 │   ├── memory/           # Memory-provider plugins (honcho, mem0, supermemory, ...)
 │   ├── context_engine/   # Context-engine plugins
+│   ├── model-providers/  # Inference backend plugins (openrouter, anthropic, gmi, ...)
 │   ├── kanban/           # Multi-agent board dispatcher + worker plugin
 │   ├── hermes-achievements/  # Gamified achievement tracking
 │   ├── observability/    # Metrics / traces / logs plugin
@@ -512,6 +513,31 @@ generic plugin surface (new hook, new ctx method) — never hardcode
 plugin-specific logic into core. PR #5295 removed 95 lines of hardcoded
 honcho argparse from `main.py` for exactly this reason.
 
+### Model-provider plugins (`plugins/model-providers/<name>/`)
+
+Every inference backend (openrouter, anthropic, gmi, deepseek, nvidia, …)
+ships as a plugin here. Each plugin's `__init__.py` calls
+`providers.register_provider(ProviderProfile(...))` at module load.
+`providers/__init__.py._discover_providers()` is a **lazy, separate
+discovery system** — scanned on first `get_provider_profile()` or
+`list_providers()` call, NOT by the general PluginManager.
+
+Scan order:
+1. Bundled: `<repo>/plugins/model-providers/<name>/`
+2. User: `$HERMES_HOME/plugins/model-providers/<name>/`
+3. Legacy: `<repo>/providers/<name>.py` (back-compat)
+
+User plugins of the same name override bundled ones — `register_provider()`
+is last-writer-wins. This lets third parties swap out any built-in
+profile without a repo patch.
+
+The general PluginManager records `kind: model-provider` manifests but does
+NOT import them (would double-instantiate `ProviderProfile`). Plugins
+without an explicit `kind:` get auto-coerced via a source-text heuristic
+(`register_provider` + `ProviderProfile` in `__init__.py`).
+
+Full authoring guide: `website/docs/developer-guide/model-provider-plugin.md`.
+
 ### Dashboard / context-engine / image-gen plugin directories
 
 `plugins/context_engine/`, `plugins/image_gen/`, `plugins/example-dashboard/`,

agent/auxiliary_client.py

@@ -216,7 +216,26 @@ def _fixed_temperature_for_model(
     return None
 
 # Default auxiliary models for direct API-key providers (cheap/fast for side tasks)
-_API_KEY_PROVIDER_AUX_MODELS: Dict[str, str] = {
+def _get_aux_model_for_provider(provider_id: str) -> str:
+    """Return the cheap auxiliary model for a provider.
+
+    Reads from ProviderProfile.default_aux_model first, falling back to the
+    legacy hardcoded dict for providers that predate the profiles system.
+    """
+    try:
+        from providers import get_provider_profile
+        _p = get_provider_profile(provider_id)
+        if _p and _p.default_aux_model:
+            return _p.default_aux_model
+    except Exception:
+        pass
+    return _API_KEY_PROVIDER_AUX_MODELS_FALLBACK.get(provider_id, "")
+
+
+# Fallback for providers not yet migrated to ProviderProfile.default_aux_model,
+# plus providers we intentionally keep pinned here (e.g. Anthropic predates
+# profiles). New providers should set default_aux_model on their profile instead.
+_API_KEY_PROVIDER_AUX_MODELS_FALLBACK: Dict[str, str] = {
     "gemini": "gemini-3-flash-preview",
     "zai": "glm-4.5-flash",
     "kimi-coding": "kimi-k2-turbo-preview",
@@ -235,6 +254,10 @@ _API_KEY_PROVIDER_AUX_MODELS: Dict[str, str] = {
     "tencent-tokenhub": "hy3-preview",
 }
 
+# Legacy alias — callers that haven't been updated to _get_aux_model_for_provider()
+# can still use this dict directly. Kept in sync with _FALLBACK above.
+_API_KEY_PROVIDER_AUX_MODELS: Dict[str, str] = _API_KEY_PROVIDER_AUX_MODELS_FALLBACK
+
 # Vision-specific model overrides for direct providers.
 # When the user's main provider has a dedicated vision/multimodal model that
 # differs from their main chat model, map it here.  The vision auto-detect
@@ -1155,7 +1178,7 @@ def _resolve_api_key_provider() -> Tuple[Optional[OpenAI], Optional[str]]:
 
             raw_base_url = _pool_runtime_base_url(entry, pconfig.inference_base_url) or pconfig.inference_base_url
             base_url = _to_openai_base_url(raw_base_url)
-            model = _API_KEY_PROVIDER_AUX_MODELS.get(provider_id)
+            model = _get_aux_model_for_provider(provider_id) or None
             if model is None:
                 continue  # skip provider if we don't know a valid aux model
             logger.debug("Auxiliary text client: %s (%s) via pool", pconfig.name, model)
@@ -1171,6 +1194,14 @@ def _resolve_api_key_provider() -> Tuple[Optional[OpenAI], Optional[str]]:
                 from hermes_cli.models import copilot_default_headers
 
                 extra["default_headers"] = copilot_default_headers()
+            else:
+                try:
+                    from providers import get_provider_profile as _gpf_aux
+                    _ph_aux = _gpf_aux(provider_id)
+                    if _ph_aux and _ph_aux.default_headers:
+                        extra["default_headers"] = dict(_ph_aux.default_headers)
+                except Exception:
+                    pass
             _client = OpenAI(api_key=api_key, base_url=base_url, **extra)
             _client = _maybe_wrap_anthropic(_client, model, api_key, raw_base_url)
             return _client, model
@@ -1182,7 +1213,7 @@ def _resolve_api_key_provider() -> Tuple[Optional[OpenAI], Optional[str]]:
 
         raw_base_url = str(creds.get("base_url", "")).strip().rstrip("/") or pconfig.inference_base_url
         base_url = _to_openai_base_url(raw_base_url)
-        model = _API_KEY_PROVIDER_AUX_MODELS.get(provider_id)
+        model = _get_aux_model_for_provider(provider_id) or None
         if model is None:
             continue  # skip provider if we don't know a valid aux model
         logger.debug("Auxiliary text client: %s (%s)", pconfig.name, model)
@@ -1198,6 +1229,14 @@ def _resolve_api_key_provider() -> Tuple[Optional[OpenAI], Optional[str]]:
             from hermes_cli.models import copilot_default_headers
 
             extra["default_headers"] = copilot_default_headers()
+        else:
+            try:
+                from providers import get_provider_profile as _gpf_aux2
+                _ph_aux2 = _gpf_aux2(provider_id)
+                if _ph_aux2 and _ph_aux2.default_headers:
+                    extra["default_headers"] = dict(_ph_aux2.default_headers)
+            except Exception:
+                pass
         _client = OpenAI(api_key=api_key, base_url=base_url, **extra)
         _client = _maybe_wrap_anthropic(_client, model, api_key, raw_base_url)
         return _client, model
@@ -1570,7 +1609,7 @@ def _try_anthropic(explicit_api_key: str = None) -> Tuple[Optional[Any], Optiona
 
     from agent.anthropic_adapter import _is_oauth_token
     is_oauth = _is_oauth_token(token)
-    model = _API_KEY_PROVIDER_AUX_MODELS.get("anthropic", "claude-haiku-4-5-20251001")
+    model = _get_aux_model_for_provider("anthropic") or "claude-haiku-4-5-20251001"
     logger.debug("Auxiliary client: Anthropic native (%s) at %s (oauth=%s)", model, base_url, is_oauth)
     try:
         real_client = build_anthropic_client(token, base_url)
@@ -2373,7 +2412,7 @@ def resolve_provider_client(
         if explicit_base_url:
             base_url = _to_openai_base_url(explicit_base_url.strip().rstrip("/"))
 
-        default_model = _API_KEY_PROVIDER_AUX_MODELS.get(provider, "")
+        default_model = _get_aux_model_for_provider(provider)
         final_model = _normalize_resolved_model(model or default_model, provider)
 
         if provider == "gemini":

agent/model_metadata.py

@@ -318,6 +318,17 @@ _URL_TO_PROVIDER: Dict[str, str] = {
     "ollama.com": "ollama-cloud",
 }
 
+# Auto-extend with hostnames derived from provider profiles.
+# Any provider with a base_url not already in the map gets added automatically.
+try:
+    from providers import list_providers as _list_providers
+    for _pp in _list_providers():
+        _host = _pp.get_hostname()
+        if _host and _host not in _URL_TO_PROVIDER:
+            _URL_TO_PROVIDER[_host] = _pp.name
+except Exception:
+    pass
+
 
 def _infer_provider_from_url(base_url: str) -> Optional[str]:
     """Infer the models.dev provider name from a base URL.

agent/transports/__init__.py

@@ -6,9 +6,16 @@ Usage:
     result = transport.normalize_response(raw_response)
 """
 
-from agent.transports.types import NormalizedResponse, ToolCall, Usage, build_tool_call, map_finish_reason  # noqa: F401
+from agent.transports.types import (
+    NormalizedResponse,
+    ToolCall,
+    Usage,
+    build_tool_call,
+    map_finish_reason,
+)  # noqa: F401
 
 _REGISTRY: dict = {}
+_discovered: bool = False
 
 
 def register_transport(api_mode: str, transport_cls: type) -> None:
@@ -23,6 +30,9 @@ def get_transport(api_mode: str):
     This allows gradual migration — call sites can check for None
     and fall back to the legacy code path.
     """
+    global _discovered
+    if not _discovered:
+        _discover_transports()
     cls = _REGISTRY.get(api_mode)
     if cls is None:
         # The registry can be partially populated when a specific transport
@@ -38,6 +48,8 @@ def get_transport(api_mode: str):
 
 def _discover_transports() -> None:
     """Import all transport modules to trigger auto-registration."""
+    global _discovered
+    _discovered = True
     try:
         import agent.transports.anthropic  # noqa: F401
     except ImportError:

agent/transports/chat_completions.py

@@ -109,7 +109,9 @@ class ChatCompletionsTransport(ProviderTransport):
     def api_mode(self) -> str:
         return "chat_completions"
 
-    def convert_messages(self, messages: List[Dict[str, Any]], **kwargs) -> List[Dict[str, Any]]:
+    def convert_messages(
+        self, messages: list[dict[str, Any]], **kwargs
+    ) -> list[dict[str, Any]]:
         """Messages are already in OpenAI format — sanitize Codex leaks only.
 
         Strips Codex Responses API fields (``codex_reasoning_items`` /
@@ -126,7 +128,9 @@ class ChatCompletionsTransport(ProviderTransport):
             tool_calls = msg.get("tool_calls")
             if isinstance(tool_calls, list):
                 for tc in tool_calls:
-                    if isinstance(tc, dict) and ("call_id" in tc or "response_item_id" in tc):
+                    if isinstance(tc, dict) and (
+                        "call_id" in tc or "response_item_id" in tc
+                    ):
                         needs_sanitize = True
                         break
                 if needs_sanitize:
@@ -149,39 +153,41 @@ class ChatCompletionsTransport(ProviderTransport):
                         tc.pop("response_item_id", None)
         return sanitized
 
-    def convert_tools(self, tools: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    def convert_tools(self, tools: list[dict[str, Any]]) -> list[dict[str, Any]]:
         """Tools are already in OpenAI format — identity."""
         return tools
 
     def build_kwargs(
         self,
         model: str,
-        messages: List[Dict[str, Any]],
-        tools: Optional[List[Dict[str, Any]]] = None,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]] | None = None,
         **params,
-    ) -> Dict[str, Any]:
+    ) -> dict[str, Any]:
         """Build chat.completions.create() kwargs.
 
-        This is the most complex transport method — it handles ~16 providers
-        via params rather than subclasses.
-
-        params:
+        params (all optional):
             timeout: float — API call timeout
             max_tokens: int | None — user-configured max tokens
-            ephemeral_max_output_tokens: int | None — one-shot override (error recovery)
+            ephemeral_max_output_tokens: int | None — one-shot override
             max_tokens_param_fn: callable — returns {max_tokens: N} or {max_completion_tokens: N}
             reasoning_config: dict | None
             request_overrides: dict | None
             session_id: str | None
-            qwen_session_metadata: dict | None — {sessionId, promptId} precomputed
             model_lower: str — lowercase model name for pattern matching
-            # Provider detection flags (all optional, default False)
+            # Provider profile path (all per-provider quirks live in providers/)
+            provider_profile: ProviderProfile | None — when present, delegates to
+                _build_kwargs_from_profile(); all flag params below are bypassed.
+            # Legacy-path flags — only used when provider_profile is None
+            # (i.e. custom / unregistered providers). Known providers all go
+            # through provider_profile.
             is_openrouter: bool
             is_nous: bool
             is_qwen_portal: bool
             is_github_models: bool
             is_nvidia_nim: bool
             is_kimi: bool
+            is_tokenhub: bool
             is_lmstudio: bool
             is_custom_provider: bool
             ollama_num_ctx: int | None
@@ -190,6 +196,7 @@ class ChatCompletionsTransport(ProviderTransport):
             # Qwen-specific
             qwen_prepare_fn: callable | None — runs AFTER codex sanitization
             qwen_prepare_inplace_fn: callable | None — in-place variant for deepcopied lists
+            qwen_session_metadata: dict | None
             # Temperature
             fixed_temperature: Any — from _fixed_temperature_for_model()
             omit_temperature: bool
@@ -199,28 +206,21 @@ class ChatCompletionsTransport(ProviderTransport):
             lmstudio_reasoning_options: list[str] | None  # raw allowed_options from /api/v1/models
             # Claude on OpenRouter/Nous max output
             anthropic_max_output: int | None
-            # Extra
-            extra_body_additions: dict | None — pre-built extra_body entries
+            extra_body_additions: dict | None
         """
         # Codex sanitization: drop reasoning_items / call_id / response_item_id
         sanitized = self.convert_messages(messages)
 
-        # Qwen portal prep AFTER codex sanitization.  If sanitize already
-        # deepcopied, reuse that copy via the in-place variant to avoid a
-        # second deepcopy.
-        is_qwen = params.get("is_qwen_portal", False)
-        if is_qwen:
-            qwen_prep = params.get("qwen_prepare_fn")
-            qwen_prep_inplace = params.get("qwen_prepare_inplace_fn")
-            if sanitized is messages:
-                if qwen_prep is not None:
-                    sanitized = qwen_prep(sanitized)
-            else:
-                # Already deepcopied — transform in place
-                if qwen_prep_inplace is not None:
-                    qwen_prep_inplace(sanitized)
-                elif qwen_prep is not None:
-                    sanitized = qwen_prep(sanitized)
+        # ── Provider profile: single-path when present ──────────────────
+        _profile = params.get("provider_profile")
+        if _profile:
+            return self._build_kwargs_from_profile(
+                _profile, model, sanitized, tools, params
+            )
+
+        # ── Legacy fallback (unregistered / unknown provider) ───────────
+        # Reached only when get_provider_profile() returned None.
+        # Known providers always go through the profile path above.
 
         # Developer role swap for GPT-5/Codex models
         model_lower = params.get("model_lower", (model or "").lower())
@@ -233,7 +233,7 @@ class ChatCompletionsTransport(ProviderTransport):
             sanitized = list(sanitized)
             sanitized[0] = {**sanitized[0], "role": "developer"}
 
-        api_kwargs: Dict[str, Any] = {
+        api_kwargs: dict[str, Any] = {
             "model": model,
             "messages": sanitized,
         }
@@ -242,19 +242,6 @@ class ChatCompletionsTransport(ProviderTransport):
         if timeout is not None:
             api_kwargs["timeout"] = timeout
 
-        # Temperature
-        fixed_temp = params.get("fixed_temperature")
-        omit_temp = params.get("omit_temperature", False)
-        if omit_temp:
-            api_kwargs.pop("temperature", None)
-        elif fixed_temp is not None:
-            api_kwargs["temperature"] = fixed_temp
-
-        # Qwen metadata (caller precomputes {sessionId, promptId})
-        qwen_meta = params.get("qwen_session_metadata")
-        if qwen_meta and is_qwen:
-            api_kwargs["metadata"] = qwen_meta
-
         # Tools
         if tools:
             # Moonshot/Kimi uses a stricter flavored JSON Schema.  Rewriting
@@ -278,13 +265,6 @@ class ChatCompletionsTransport(ProviderTransport):
             api_kwargs.update(max_tokens_fn(ephemeral))
         elif max_tokens is not None and max_tokens_fn:
             api_kwargs.update(max_tokens_fn(max_tokens))
-        elif is_nvidia_nim and max_tokens_fn:
-            api_kwargs.update(max_tokens_fn(16384))
-        elif is_qwen and max_tokens_fn:
-            api_kwargs.update(max_tokens_fn(65536))
-        elif is_kimi and max_tokens_fn:
-            # Kimi/Moonshot: 32000 matches Kimi CLI's default
-            api_kwargs.update(max_tokens_fn(32000))
         elif anthropic_max_out is not None:
             api_kwargs["max_tokens"] = anthropic_max_out
 
@@ -331,7 +311,7 @@ class ChatCompletionsTransport(ProviderTransport):
                 api_kwargs["reasoning_effort"] = _lm_effort
 
         # extra_body assembly
-        extra_body: Dict[str, Any] = {}
+        extra_body: dict[str, Any] = {}
 
         is_openrouter = params.get("is_openrouter", False)
         is_nous = params.get("is_nous", False)
@@ -361,35 +341,7 @@ class ChatCompletionsTransport(ProviderTransport):
                 if gh_reasoning is not None:
                     extra_body["reasoning"] = gh_reasoning
             else:
-                if reasoning_config is not None:
-                    rc = dict(reasoning_config)
-                    if is_nous and rc.get("enabled") is False:
-                        pass  # omit for Nous when disabled
-                    else:
-                        extra_body["reasoning"] = rc
-                else:
-                    extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
-
-        if is_nous:
-            extra_body["tags"] = ["product=hermes-agent"]
-
-        # Ollama num_ctx
-        ollama_ctx = params.get("ollama_num_ctx")
-        if ollama_ctx:
-            options = extra_body.get("options", {})
-            options["num_ctx"] = ollama_ctx
-            extra_body["options"] = options
-
-        # Ollama/custom think=false
-        if params.get("is_custom_provider", False):
-            if reasoning_config and isinstance(reasoning_config, dict):
-                _effort = (reasoning_config.get("effort") or "").strip().lower()
-                _enabled = reasoning_config.get("enabled", True)
-                if _effort == "none" or _enabled is False:
-                    extra_body["think"] = False
-
-        if is_qwen:
-            extra_body["vl_high_resolution_images"] = True
+                extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
 
         if provider_name == "gemini":
             raw_thinking_config = _build_gemini_thinking_config(model, reasoning_config)
@@ -423,6 +375,120 @@ class ChatCompletionsTransport(ProviderTransport):
 
         return api_kwargs
 
+    def _build_kwargs_from_profile(self, profile, model, sanitized, tools, params):
+        """Build API kwargs using a ProviderProfile — single path, no legacy flags.
+
+        This method replaces the entire flag-based kwargs assembly when a
+        provider_profile is passed. Every quirk comes from the profile object.
+        """
+        from providers.base import OMIT_TEMPERATURE
+
+        # Message preprocessing
+        sanitized = profile.prepare_messages(sanitized)
+
+        # Developer role swap — model-name-based, applies to all providers
+        _model_lower = (model or "").lower()
+        if (
+            sanitized
+            and isinstance(sanitized[0], dict)
+            and sanitized[0].get("role") == "system"
+            and any(p in _model_lower for p in DEVELOPER_ROLE_MODELS)
+        ):
+            sanitized = list(sanitized)
+            sanitized[0] = {**sanitized[0], "role": "developer"}
+
+        api_kwargs: dict[str, Any] = {
+            "model": model,
+            "messages": sanitized,
+        }
+
+        # Temperature
+        if profile.fixed_temperature is OMIT_TEMPERATURE:
+            pass  # Don't include temperature at all
+        elif profile.fixed_temperature is not None:
+            api_kwargs["temperature"] = profile.fixed_temperature
+        else:
+            # Use caller's temperature if provided
+            temp = params.get("temperature")
+            if temp is not None:
+                api_kwargs["temperature"] = temp
+
+        # Timeout
+        timeout = params.get("timeout")
+        if timeout is not None:
+            api_kwargs["timeout"] = timeout
+
+        # Tools — apply Moonshot/Kimi schema sanitization regardless of path
+        if tools:
+            if is_moonshot_model(model):
+                tools = sanitize_moonshot_tools(tools)
+            api_kwargs["tools"] = tools
+
+        # max_tokens resolution — priority: ephemeral > user > profile default
+        max_tokens_fn = params.get("max_tokens_param_fn")
+        ephemeral = params.get("ephemeral_max_output_tokens")
+        user_max = params.get("max_tokens")
+        anthropic_max = params.get("anthropic_max_output")
+
+        if ephemeral is not None and max_tokens_fn:
+            api_kwargs.update(max_tokens_fn(ephemeral))
+        elif user_max is not None and max_tokens_fn:
+            api_kwargs.update(max_tokens_fn(user_max))
+        elif profile.default_max_tokens and max_tokens_fn:
+            api_kwargs.update(max_tokens_fn(profile.default_max_tokens))
+        elif anthropic_max is not None:
+            api_kwargs["max_tokens"] = anthropic_max
+
+        # Provider-specific api_kwargs extras (reasoning_effort, metadata, etc.)
+        reasoning_config = params.get("reasoning_config")
+        extra_body_from_profile, top_level_from_profile = (
+            profile.build_api_kwargs_extras(
+                reasoning_config=reasoning_config,
+                supports_reasoning=params.get("supports_reasoning", False),
+                qwen_session_metadata=params.get("qwen_session_metadata"),
+                model=model,
+                ollama_num_ctx=params.get("ollama_num_ctx"),
+            )
+        )
+        api_kwargs.update(top_level_from_profile)
+
+        # extra_body assembly
+        extra_body: dict[str, Any] = {}
+
+        # Profile's extra_body (tags, provider prefs, vl_high_resolution, etc.)
+        profile_body = profile.build_extra_body(
+            session_id=params.get("session_id"),
+            provider_preferences=params.get("provider_preferences"),
+            model=model,
+            base_url=params.get("base_url"),
+            reasoning_config=reasoning_config,
+        )
+        if profile_body:
+            extra_body.update(profile_body)
+
+        # Profile's reasoning/thinking extra_body entries
+        if extra_body_from_profile:
+            extra_body.update(extra_body_from_profile)
+
+        # Merge any pre-built extra_body additions from the caller
+        additions = params.get("extra_body_additions")
+        if additions:
+            extra_body.update(additions)
+
+        # Request overrides (user config)
+        overrides = params.get("request_overrides")
+        if overrides:
+            for k, v in overrides.items():
+                if k == "extra_body" and isinstance(v, dict):
+                    extra_body.update(v)
+                else:
+                    api_kwargs[k] = v
+
+        if extra_body:
+            api_kwargs["extra_body"] = extra_body
+
+        return api_kwargs
+
     def normalize_response(self, response: Any, **kwargs) -> NormalizedResponse:
         """Normalize OpenAI ChatCompletion to NormalizedResponse.
 
@@ -444,7 +510,7 @@ class ChatCompletionsTransport(ProviderTransport):
                 # Gemini 3 thinking models attach extra_content with
                 # thought_signature — without replay on the next turn the API
                 # rejects the request with 400.
-                tc_provider_data: Dict[str, Any] = {}
+                tc_provider_data: dict[str, Any] = {}
                 extra = getattr(tc, "extra_content", None)
                 if extra is None and hasattr(tc, "model_extra"):
                     extra = (tc.model_extra or {}).get("extra_content")
@@ -455,12 +521,14 @@ class ChatCompletionsTransport(ProviderTransport):
                         except Exception:
                             pass
                     tc_provider_data["extra_content"] = extra
-                tool_calls.append(ToolCall(
-                    id=tc.id,
-                    name=tc.function.name,
-                    arguments=tc.function.arguments,
-                    provider_data=tc_provider_data or None,
-                ))
+                tool_calls.append(
+                    ToolCall(
+                        id=tc.id,
+                        name=tc.function.name,
+                        arguments=tc.function.arguments,
+                        provider_data=tc_provider_data or None,
+                    )
+                )
 
         usage = None
         if hasattr(response, "usage") and response.usage:
@@ -508,7 +576,7 @@ class ChatCompletionsTransport(ProviderTransport):
             return False
         return True
 
-    def extract_cache_stats(self, response: Any) -> Optional[Dict[str, int]]:
+    def extract_cache_stats(self, response: Any) -> dict[str, int] | None:
         """Extract OpenRouter/OpenAI cache stats from prompt_tokens_details."""
         usage = getattr(response, "usage", None)
         if usage is None:

agent/transports/types.py

@@ -12,7 +12,7 @@ from __future__ import annotations
 
 import json
 from dataclasses import dataclass, field
-from typing import Any, Dict, List, Optional
+from typing import Any
 
 
 @dataclass
@@ -32,10 +32,10 @@ class ToolCall:
     * Others: ``None``
     """
 
-    id: Optional[str]
+    id: str | None
     name: str
     arguments: str  # JSON string
-    provider_data: Optional[Dict[str, Any]] = field(default=None, repr=False)
+    provider_data: dict[str, Any] | None = field(default=None, repr=False)
 
     # ── Backward compatibility ──────────────────────────────────
     # The agent loop reads tc.function.name / tc.function.arguments
@@ -47,17 +47,17 @@ class ToolCall:
         return "function"
 
     @property
-    def function(self) -> "ToolCall":
+    def function(self) -> ToolCall:
         """Return self so tc.function.name / tc.function.arguments work."""
         return self
 
     @property
-    def call_id(self) -> Optional[str]:
+    def call_id(self) -> str | None:
         """Codex call_id from provider_data, accessed via getattr by _build_assistant_message."""
         return (self.provider_data or {}).get("call_id")
 
     @property
-    def response_item_id(self) -> Optional[str]:
+    def response_item_id(self) -> str | None:
         """Codex response_item_id from provider_data."""
         return (self.provider_data or {}).get("response_item_id")
 
@@ -101,18 +101,18 @@ class NormalizedResponse:
     * Others: ``None``
     """
 
-    content: Optional[str]
-    tool_calls: Optional[List[ToolCall]]
+    content: str | None
+    tool_calls: list[ToolCall] | None
     finish_reason: str  # "stop", "tool_calls", "length", "content_filter"
-    reasoning: Optional[str] = None
-    usage: Optional[Usage] = None
-    provider_data: Optional[Dict[str, Any]] = field(default=None, repr=False)
+    reasoning: str | None = None
+    usage: Usage | None = None
+    provider_data: dict[str, Any] | None = field(default=None, repr=False)
 
     # ── Backward compatibility ──────────────────────────────────
     # The shim _nr_to_assistant_message() mapped these from provider_data.
     # These properties let NormalizedResponse pass through directly.
     @property
-    def reasoning_content(self) -> Optional[str]:
+    def reasoning_content(self) -> str | None:
         pd = self.provider_data or {}
         return pd.get("reasoning_content")
 
@@ -136,8 +136,9 @@ class NormalizedResponse:
 # Factory helpers
 # ---------------------------------------------------------------------------
 
+
 def build_tool_call(
-    id: Optional[str],
+    id: str | None,
     name: str,
     arguments: Any,
     **provider_fields: Any,
@@ -151,7 +152,7 @@ def build_tool_call(
     return ToolCall(id=id, name=name, arguments=args_str, provider_data=pd)
 
 
-def map_finish_reason(reason: Optional[str], mapping: Dict[str, str]) -> str:
+def map_finish_reason(reason: str | None, mapping: dict[str, str]) -> str:
     """Translate a provider-specific stop reason to the normalised set.
 
     Falls back to ``"stop"`` for unknown or ``None`` reasons.

hermes_cli/auth.py

@@ -416,6 +416,40 @@ PROVIDER_REGISTRY: Dict[str, ProviderConfig] = {
     ),
 }
 
+# Auto-extend PROVIDER_REGISTRY with any api-key provider registered in
+# providers/ that is not already declared above.  New providers only need a
+# plugins/model-providers/<name>/ plugin — no edits to this file required.
+try:
+    from providers import list_providers as _list_providers_for_registry
+    for _pp in _list_providers_for_registry():
+        if _pp.name in PROVIDER_REGISTRY:
+            continue
+        if _pp.auth_type != "api_key" or not _pp.env_vars:
+            continue
+        # Skip providers that need custom token resolution or are special-cased
+        # in resolve_provider() (copilot/kimi/zai have bespoke token refresh;
+        # openrouter/custom are aggregator/user-supplied and handled outside
+        # the registry — adding them here breaks runtime_provider resolution
+        # that relies on `openrouter not in PROVIDER_REGISTRY`).
+        if _pp.name in {"copilot", "kimi-coding", "kimi-coding-cn", "zai", "openrouter", "custom"}:
+            continue
+        _api_key_vars = tuple(v for v in _pp.env_vars if not v.endswith("_BASE_URL") and not v.endswith("_URL"))
+        _base_url_var = next((v for v in _pp.env_vars if v.endswith("_BASE_URL") or v.endswith("_URL")), None)
+        PROVIDER_REGISTRY[_pp.name] = ProviderConfig(
+            id=_pp.name,
+            name=_pp.display_name or _pp.name,
+            auth_type="api_key",
+            inference_base_url=_pp.base_url,
+            api_key_env_vars=_api_key_vars or _pp.env_vars,
+            base_url_env_var=_base_url_var or "",
+        )
+        # Also register aliases so resolve_provider() resolves them
+        for _alias in _pp.aliases:
+            if _alias not in PROVIDER_REGISTRY:
+                PROVIDER_REGISTRY[_alias] = PROVIDER_REGISTRY[_pp.name]
+except Exception:
+    pass
+
 
 # =============================================================================
 # Anthropic Key Helper
@@ -1195,6 +1229,17 @@ def resolve_provider(
         "vllm": "custom", "llamacpp": "custom",
         "llama.cpp": "custom", "llama-cpp": "custom",
     }
+    # Extend with aliases declared in plugins/model-providers/<name>/ that aren't already mapped.
+    # This keeps providers/ as the single source for new aliases while the
+    # hardcoded dict above remains authoritative for existing ones.
+    try:
+        from providers import list_providers as _lp
+        for _pp in _lp():
+            for _alias in _pp.aliases:
+                if _alias not in _PROVIDER_ALIASES:
+                    _PROVIDER_ALIASES[_alias] = _pp.name
+    except Exception:
+        pass
     normalized = _PROVIDER_ALIASES.get(normalized, normalized)
 
     if normalized == "openrouter":

hermes_cli/config.py

@@ -4840,3 +4840,45 @@ def config_command(args):
         print("  hermes config path      Show config file path")
         print("  hermes config env-path  Show .env file path")
         sys.exit(1)
+
+
+# ── Profile-driven env var injection ─────────────────────────────────────────
+# Any provider registered in providers/ with auth_type="api_key" automatically
+# gets its env_vars exposed in OPTIONAL_ENV_VARS without editing this file.
+# Runs once at import time.
+
+_profile_env_vars_injected = False
+
+
+def _inject_profile_env_vars() -> None:
+    """Populate OPTIONAL_ENV_VARS from provider profiles not already listed.
+
+    Called once at module load time. Idempotent — repeated calls are no-ops.
+    """
+    global _profile_env_vars_injected
+    if _profile_env_vars_injected:
+        return
+    _profile_env_vars_injected = True
+    try:
+        from providers import list_providers
+        for _pp in list_providers():
+            if _pp.auth_type not in ("api_key",):
+                continue
+            for _var in _pp.env_vars:
+                if _var in OPTIONAL_ENV_VARS:
+                    continue
+                _is_key = not _var.endswith("_BASE_URL") and not _var.endswith("_URL")
+                OPTIONAL_ENV_VARS[_var] = {
+                    "description": f"{_pp.display_name or _pp.name} {'API key' if _is_key else 'base URL override'}",
+                    "prompt": f"{_pp.display_name or _pp.name} {'API key' if _is_key else 'base URL (leave empty for default)'}",
+                    "url": _pp.signup_url or None,
+                    "password": _is_key,
+                    "category": "provider",
+                    "advanced": True,
+                }
+    except Exception:
+        pass
+
+
+# Eagerly inject so that OPTIONAL_ENV_VARS is fully populated at import time.
+_inject_profile_env_vars()

hermes_cli/doctor.py

@@ -175,6 +175,85 @@ def _check_gateway_service_linger(issues: list[str]) -> None:
         check_warn("Could not verify systemd linger", f"({linger_detail})")
 
 
+_APIKEY_PROVIDERS_CACHE: list | None = None
+
+
+def _build_apikey_providers_list() -> list:
+    """Build the API-key provider health-check list once and cache it.
+
+    Tuple format: (name, env_vars, default_url, base_env, supports_models_endpoint)
+    Base list augmented with any ProviderProfile with auth_type="api_key" not
+    already present — adding plugins/model-providers/<name>/ is sufficient to get into doctor.
+    """
+    _static = [
+        ("Z.AI / GLM",      ("GLM_API_KEY", "ZAI_API_KEY", "Z_AI_API_KEY"), "https://api.z.ai/api/paas/v4/models", "GLM_BASE_URL", True),
+        ("Kimi / Moonshot",  ("KIMI_API_KEY",),                              "https://api.moonshot.ai/v1/models",   "KIMI_BASE_URL", True),
+        ("StepFun Step Plan", ("STEPFUN_API_KEY",),                          "https://api.stepfun.ai/step_plan/v1/models", "STEPFUN_BASE_URL", True),
+        ("Kimi / Moonshot (China)", ("KIMI_CN_API_KEY",),                    "https://api.moonshot.cn/v1/models",   None, True),
+        ("Arcee AI",         ("ARCEEAI_API_KEY",),                           "https://api.arcee.ai/api/v1/models",  "ARCEE_BASE_URL", True),
+        ("GMI Cloud",        ("GMI_API_KEY",),                               "https://api.gmi-serving.com/v1/models", "GMI_BASE_URL", True),
+        ("DeepSeek",         ("DEEPSEEK_API_KEY",),                          "https://api.deepseek.com/v1/models",  "DEEPSEEK_BASE_URL", True),
+        ("Hugging Face",     ("HF_TOKEN",),                                  "https://router.huggingface.co/v1/models", "HF_BASE_URL", True),
+        ("NVIDIA NIM",       ("NVIDIA_API_KEY",),                            "https://integrate.api.nvidia.com/v1/models", "NVIDIA_BASE_URL", True),
+        ("Alibaba/DashScope", ("DASHSCOPE_API_KEY",),                        "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/models", "DASHSCOPE_BASE_URL", True),
+        # MiniMax global: /v1 endpoint supports /models.
+        ("MiniMax",          ("MINIMAX_API_KEY",),                           "https://api.minimax.io/v1/models",    "MINIMAX_BASE_URL", True),
+        # MiniMax CN: /v1 endpoint does NOT support /models (returns 404).
+        ("MiniMax (China)",  ("MINIMAX_CN_API_KEY",),                        "https://api.minimaxi.com/v1/models",  "MINIMAX_CN_BASE_URL", False),
+        ("Vercel AI Gateway", ("AI_GATEWAY_API_KEY",),                       "https://ai-gateway.vercel.sh/v1/models", "AI_GATEWAY_BASE_URL", True),
+        ("Kilo Code",        ("KILOCODE_API_KEY",),                          "https://api.kilo.ai/api/gateway/models", "KILOCODE_BASE_URL", True),
+        ("OpenCode Zen",     ("OPENCODE_ZEN_API_KEY",),                      "https://opencode.ai/zen/v1/models",  "OPENCODE_ZEN_BASE_URL", True),
+        # OpenCode Go has no shared /models endpoint; skip the health check.
+        ("OpenCode Go",      ("OPENCODE_GO_API_KEY",),                       None,                                  "OPENCODE_GO_BASE_URL", False),
+    ]
+    _known_names = {t[0] for t in _static}
+    # Also index by profile canonical name so profiles without display_name
+    # don't create duplicate entries for providers already in the static list.
+    _known_canonical: set[str] = set()
+    _name_to_canonical = {
+        "Z.AI / GLM": "zai", "Kimi / Moonshot": "kimi-coding",
+        "StepFun Step Plan": "stepfun", "Kimi / Moonshot (China)": "kimi-coding-cn",
+        "Arcee AI": "arcee", "GMI Cloud": "gmi", "DeepSeek": "deepseek",
+        "Hugging Face": "huggingface", "NVIDIA NIM": "nvidia",
+        "Alibaba/DashScope": "alibaba", "MiniMax": "minimax",
+        "MiniMax (China)": "minimax-cn", "Vercel AI Gateway": "ai-gateway",
+        "Kilo Code": "kilocode", "OpenCode Zen": "opencode-zen",
+        "OpenCode Go": "opencode-go",
+    }
+    for _label, _canonical in _name_to_canonical.items():
+        _known_canonical.add(_canonical)
+    try:
+        from providers import list_providers
+        from providers.base import ProviderProfile as _PP
+        for _pp in list_providers():
+            if not isinstance(_pp, _PP) or _pp.auth_type != "api_key" or not _pp.env_vars:
+                continue
+            _label = _pp.display_name or _pp.name
+            if _label in _known_names or _pp.name in _known_canonical:
+                continue
+            # Separate API-key vars from base-URL override vars — the health-check
+            # loop sends the first found value as Authorization: Bearer, so a URL
+            # string must never be picked.
+            _key_vars = tuple(
+                v for v in _pp.env_vars
+                if not v.endswith("_BASE_URL") and not v.endswith("_URL")
+            )
+            _base_var = next(
+                (v for v in _pp.env_vars if v.endswith("_BASE_URL") or v.endswith("_URL")),
+                None,
+            )
+            if not _key_vars:
+                continue
+            _models_url = (
+                (_pp.models_url or (_pp.base_url.rstrip("/") + "/models"))
+                if _pp.base_url else None
+            )
+            _static.append((_label, _key_vars, _models_url, _base_var, True))
+    except Exception:
+        pass
+    return _static
+
+
 def run_doctor(args):
     """Run diagnostic checks."""
     should_fix = getattr(args, 'fix', False)
@@ -1087,27 +1166,11 @@ def run_doctor(args):
     # -- API-key providers --
     # Tuple: (name, env_vars, default_url, base_env, supports_models_endpoint)
     # If supports_models_endpoint is False, we skip the health check and just show "configured"
-    _apikey_providers = [
-        ("Z.AI / GLM",      ("GLM_API_KEY", "ZAI_API_KEY", "Z_AI_API_KEY"), "https://api.z.ai/api/paas/v4/models", "GLM_BASE_URL", True),
-        ("Kimi / Moonshot",  ("KIMI_API_KEY",),                              "https://api.moonshot.ai/v1/models",   "KIMI_BASE_URL", True),
-        ("StepFun Step Plan",   ("STEPFUN_API_KEY",),                           "https://api.stepfun.ai/step_plan/v1/models", "STEPFUN_BASE_URL", True),
-        ("Kimi / Moonshot (China)", ("KIMI_CN_API_KEY",),                    "https://api.moonshot.cn/v1/models",   None, True),
-        ("Arcee AI",         ("ARCEEAI_API_KEY",),                            "https://api.arcee.ai/api/v1/models",  "ARCEE_BASE_URL", True),
-        ("GMI Cloud",        ("GMI_API_KEY",),                                "https://api.gmi-serving.com/v1/models", "GMI_BASE_URL", True),
-        ("DeepSeek",         ("DEEPSEEK_API_KEY",),                           "https://api.deepseek.com/v1/models",  "DEEPSEEK_BASE_URL", True),
-        ("Hugging Face",     ("HF_TOKEN",),                                   "https://router.huggingface.co/v1/models", "HF_BASE_URL", True),
-        ("NVIDIA NIM",       ("NVIDIA_API_KEY",),                             "https://integrate.api.nvidia.com/v1/models", "NVIDIA_BASE_URL", True),
-        ("Alibaba/DashScope", ("DASHSCOPE_API_KEY",),                         "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/models", "DASHSCOPE_BASE_URL", True),
-        # MiniMax global: /v1 endpoint supports /models.
-        ("MiniMax",          ("MINIMAX_API_KEY",),                            "https://api.minimax.io/v1/models",    "MINIMAX_BASE_URL", True),
-        # MiniMax CN: /v1 endpoint does NOT support /models (returns 404).
-        ("MiniMax (China)",  ("MINIMAX_CN_API_KEY",),                         "https://api.minimaxi.com/v1/models",  "MINIMAX_CN_BASE_URL", False),
-        ("Vercel AI Gateway",       ("AI_GATEWAY_API_KEY",),                          "https://ai-gateway.vercel.sh/v1/models", "AI_GATEWAY_BASE_URL", True),
-        ("Kilo Code",        ("KILOCODE_API_KEY",),                            "https://api.kilo.ai/api/gateway/models",  "KILOCODE_BASE_URL", True),
-        ("OpenCode Zen",     ("OPENCODE_ZEN_API_KEY",),                        "https://opencode.ai/zen/v1/models",  "OPENCODE_ZEN_BASE_URL", True),
-        # OpenCode Go has no shared /models endpoint; skip the health check.
-        ("OpenCode Go",      ("OPENCODE_GO_API_KEY",),                         None,                                  "OPENCODE_GO_BASE_URL", False),
-    ]
+    # Cached at module level after first build — profiles auto-extend it.
+    global _APIKEY_PROVIDERS_CACHE
+    if _APIKEY_PROVIDERS_CACHE is None:
+        _APIKEY_PROVIDERS_CACHE = _build_apikey_providers_list()
+    _apikey_providers = _APIKEY_PROVIDERS_CACHE
     for _pname, _env_vars, _default_url, _base_env, _supports_health_check in _apikey_providers:
         _key = ""
         for _ev in _env_vars:

hermes_cli/main.py

@@ -1611,6 +1611,21 @@ def cmd_model(args):
     select_provider_and_model(args=args)
 
 
+def _is_profile_api_key_provider(provider_id: str) -> bool:
+    """Return True when provider_id maps to a profile with auth_type='api_key'.
+
+    Used as a catch-all in select_provider_and_model() so that new providers
+    declared in plugins/model-providers/<name>/ automatically dispatch to _model_flow_api_key_provider
+    without requiring an explicit elif branch here.
+    """
+    try:
+        from providers import get_provider_profile
+        _p = get_provider_profile(provider_id)
+        return _p is not None and _p.auth_type == "api_key"
+    except Exception:
+        return False
+
+
 def select_provider_and_model(args=None):
     """Core provider selection + model picking logic.
 
@@ -1907,7 +1922,7 @@ def select_provider_and_model(args=None):
         "ollama-cloud",
         "tencent-tokenhub",
         "lmstudio",
-    ):
+    ) or _is_profile_api_key_provider(selected_provider):
         _model_flow_api_key_provider(config, selected_provider, current_model)
 
     # ── Post-switch cleanup: clear stale OPENAI_BASE_URL ──────────────
@@ -8215,6 +8230,22 @@ def cmd_logs(args):
     )
 
 
+def _build_provider_choices() -> list[str]:
+    """Build the --provider choices list from CANONICAL_PROVIDERS + 'auto'."""
+    try:
+        from hermes_cli.models import CANONICAL_PROVIDERS as _cp
+        return ["auto"] + [p.slug for p in _cp]
+    except Exception:
+        # Fallback: static list guarantees the CLI always works
+        return [
+            "auto", "openrouter", "nous", "openai-codex", "copilot-acp", "copilot",
+            "anthropic", "gemini", "google-gemini-cli", "xai", "bedrock", "azure-foundry",
+            "ollama-cloud", "huggingface", "zai", "kimi-coding", "kimi-coding-cn",
+            "stepfun", "minimax", "minimax-cn", "kilocode", "xiaomi", "arcee",
+            "nvidia", "deepseek", "alibaba", "qwen-oauth", "opencode-zen", "opencode-go",
+        ]
+
+
 def main():
     """Main entry point for hermes CLI."""
     from hermes_cli._parser import build_top_level_parser

hermes_cli/models.py

@@ -806,6 +806,25 @@ CANONICAL_PROVIDERS: list[ProviderEntry] = [
     ProviderEntry("ai-gateway",     "Vercel AI Gateway",        "Vercel AI Gateway"),
 ]
 
+# Auto-extend CANONICAL_PROVIDERS with any provider registered in providers/
+# that is not already in the list above.  Adding plugins/model-providers/<name>/
+# is sufficient to expose a new provider in the model picker, /model, and all
+# downstream consumers — no edits to this file needed.
+_canonical_slugs = {p.slug for p in CANONICAL_PROVIDERS}
+try:
+    from providers import list_providers as _list_providers_for_canonical
+    for _pp in _list_providers_for_canonical():
+        if _pp.name in _canonical_slugs:
+            continue
+        if _pp.auth_type in ("oauth_device_code", "oauth_external", "external_process", "aws_sdk", "copilot"):
+            continue  # non-api-key flows need bespoke picker UX; skip auto-inject
+        _label = _pp.display_name or _pp.name
+        _desc = _pp.description or f"{_label} (direct API)"
+        CANONICAL_PROVIDERS.append(ProviderEntry(_pp.name, _label, _desc))
+        _canonical_slugs.add(_pp.name)
+except Exception:
+    pass
+
 # Derived dicts — used throughout the codebase
 _PROVIDER_LABELS = {p.slug: p.label for p in CANONICAL_PROVIDERS}
 _PROVIDER_LABELS["custom"] = "Custom endpoint"  # special case: not a named provider
@@ -2023,6 +2042,34 @@ def provider_model_ids(provider: Optional[str], *, force_refresh: bool = False)
                 return ids
         except Exception:
             pass
+
+    # ── Profile-based generic live fetch (all simple api-key providers) ──
+    # Handles any provider registered in providers/ with auth_type="api_key".
+    # Replaces per-provider copy-paste blocks (stepfun, gmi, zai, etc.).
+    try:
+        from providers import get_provider_profile
+        from hermes_cli.auth import resolve_api_key_provider_credentials
+
+        _p = get_provider_profile(normalized)
+        if _p and _p.auth_type == "api_key" and _p.base_url:
+            try:
+                creds = resolve_api_key_provider_credentials(normalized)
+                api_key = str(creds.get("api_key") or "").strip()
+                base_url = str(creds.get("base_url") or "").strip()
+            except Exception:
+                api_key, base_url = "", _p.base_url
+            if not base_url:
+                base_url = _p.base_url
+            if api_key:
+                live = _p.fetch_models(api_key=api_key)
+                if live:
+                    return live
+            # Use profile's fallback_models if defined
+            if _p.fallback_models:
+                return list(_p.fallback_models)
+    except Exception:
+        pass
+
     curated_static = list(_PROVIDER_MODELS.get(normalized, []))
     if normalized in _MODELS_DEV_PREFERRED:
         return _merge_with_models_dev(normalized, curated_static)

hermes_cli/plugins.py

@@ -173,7 +173,7 @@ def _get_enabled_plugins() -> Optional[set]:
 # Data classes
 # ---------------------------------------------------------------------------
 
-_VALID_PLUGIN_KINDS: Set[str] = {"standalone", "backend", "exclusive", "platform"}
+_VALID_PLUGIN_KINDS: Set[str] = {"standalone", "backend", "exclusive", "platform", "model-provider"}
 
 
 @dataclass
@@ -643,15 +643,17 @@ class PluginManager:
         #   - flat: ``plugins/disk-cleanup/plugin.yaml`` (standalone)
         #   - category: ``plugins/image_gen/openai/plugin.yaml`` (backend)
         #
-        # ``memory/`` and ``context_engine/`` are skipped at the top level —
-        # they have their own discovery systems. ``platforms/`` is a category
-        # holding platform adapters (scanned one level deeper below).
+        # ``memory/``, ``context_engine/``, and ``model-providers/`` are
+        # skipped at the top level — they have their own discovery systems
+        # (plugins/memory/__init__.py, providers/__init__.py). ``platforms/``
+        # is a category holding platform adapters (scanned one level deeper
+        # below).
         repo_plugins = get_bundled_plugins_dir()
         manifests.extend(
             self._scan_directory(
                 repo_plugins,
                 source="bundled",
-                skip_names={"memory", "context_engine", "platforms"},
+                skip_names={"memory", "context_engine", "platforms", "model-providers"},
             )
         )
         manifests.extend(
@@ -709,6 +711,21 @@ class PluginManager:
                 )
                 continue
 
+            # Model provider plugins are loaded by providers/__init__.py
+            # (its own lazy discovery keyed off first get_provider_profile()
+            # call). We record the manifest here for introspection but do
+            # not import the module — a second import would create two
+            # ProviderProfile instances and break the "last writer wins"
+            # override semantics between bundled and user plugins.
+            if manifest.kind == "model-provider":
+                loaded = LoadedPlugin(manifest=manifest, enabled=True)
+                self._plugins[lookup_key] = loaded
+                logger.debug(
+                    "Skipping '%s' (model-provider, handled by providers/ discovery)",
+                    lookup_key,
+                )
+                continue
+
             # Built-in backends auto-load — they ship with hermes and must
             # just work. Selection among them (e.g. which image_gen backend
             # services calls) is driven by ``<category>.provider`` config,
@@ -886,6 +903,19 @@ class PluginManager:
                                 "treating as kind='exclusive'",
                                 key,
                             )
+                        elif (
+                            "register_provider" in source_text
+                            and "ProviderProfile" in source_text
+                        ):
+                            # Model provider plugin (calls register_provider()
+                            # from ``providers`` with a ProviderProfile). Route
+                            # to providers/__init__.py discovery.
+                            kind = "model-provider"
+                            logger.debug(
+                                "Plugin %s: detected model provider, "
+                                "treating as kind='model-provider'",
+                                key,
+                            )
                     except Exception:
                         pass
 

plugins/model-providers/README.md

@@ -0,0 +1,70 @@
+# Model Provider Plugins
+
+Each subdirectory is a self-contained provider profile plugin. The
+directory layout mirrors `plugins/platforms/`:
+
+```
+plugins/model-providers/
+├── openrouter/
+│   ├── __init__.py      # registers the ProviderProfile
+│   └── plugin.yaml      # manifest: name, kind, version, description
+├── anthropic/
+│   ├── __init__.py
+│   └── plugin.yaml
+└── ...
+```
+
+## How discovery works
+
+`providers/__init__.py._discover_providers()` scans this directory (and
+`$HERMES_HOME/plugins/model-providers/`) the first time anything calls
+`get_provider_profile()` or `list_providers()`. Each `__init__.py` is
+imported and expected to call `providers.register_provider(profile)`.
+
+User plugins at `$HERMES_HOME/plugins/model-providers/<name>/` override
+bundled plugins of the same name — last-writer-wins in
+`register_provider()`. Drop a file there to replace a built-in.
+
+## Adding a new provider
+
+1. Create `plugins/model-providers/<your_provider>/__init__.py`:
+
+   ```python
+   from providers import register_provider
+   from providers.base import ProviderProfile
+
+   my_provider = ProviderProfile(
+       name="your-provider",
+       aliases=("alias1", "alias2"),
+       display_name="Your Provider",
+       description="One-line description shown in the setup picker",
+       signup_url="https://your-provider.example.com/keys",
+       env_vars=("YOUR_PROVIDER_API_KEY", "YOUR_PROVIDER_BASE_URL"),
+       base_url="https://api.your-provider.example.com/v1",
+       default_aux_model="your-cheap-model",
+   )
+
+   register_provider(my_provider)
+   ```
+
+2. Create `plugins/model-providers/<your_provider>/plugin.yaml`:
+
+   ```yaml
+   name: your-provider-profile
+   kind: model-provider
+   version: 1.0.0
+   description: Short sentence about the provider
+   author: Your Name
+   ```
+
+Nothing else needs to change. `auth.py`, `config.py`, `models.py`,
+`doctor.py`, `model_metadata.py`, `runtime_provider.py`, and the
+chat_completions transport all auto-wire from the registry.
+
+## Non-trivial profiles
+
+Override the `ProviderProfile` hooks in a subclass for per-provider
+quirks — see `plugins/model-providers/openrouter/__init__.py` for
+`build_extra_body` and `build_api_kwargs_extras` examples, and
+`plugins/model-providers/gemini/__init__.py` for `thinking_config`
+translation.

plugins/model-providers/ai-gateway/__init__.py

@@ -0,0 +1,43 @@
+"""Vercel AI Gateway provider profile.
+
+AI Gateway routes to multiple backends. Hermes sends attribution
+headers and full reasoning config passthrough.
+"""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class VercelAIGatewayProfile(ProviderProfile):
+    """Vercel AI Gateway — attribution headers + reasoning passthrough."""
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        supports_reasoning: bool = True,
+        **ctx: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        extra_body: dict[str, Any] = {}
+        if supports_reasoning and reasoning_config is not None:
+            extra_body["reasoning"] = dict(reasoning_config)
+        elif supports_reasoning:
+            extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
+        return extra_body, {}
+
+
+vercel = VercelAIGatewayProfile(
+    name="ai-gateway",
+    aliases=("vercel", "vercel-ai-gateway", "ai_gateway", "aigateway"),
+    env_vars=("AI_GATEWAY_API_KEY",),
+    base_url="https://ai-gateway.vercel.sh/v1",
+    default_headers={
+        "HTTP-Referer": "https://hermes-agent.nousresearch.com",
+        "X-Title": "Hermes Agent",
+    },
+    default_aux_model="google/gemini-3-flash",
+)
+
+register_provider(vercel)

plugins/model-providers/ai-gateway/plugin.yaml

@@ -0,0 +1,5 @@
+name: ai-gateway-provider
+kind: model-provider
+version: 1.0.0
+description: Vercel AI Gateway
+author: Nous Research

plugins/model-providers/alibaba-coding-plan/__init__.py

@@ -0,0 +1,21 @@
+"""Alibaba Cloud Coding Plan provider profile.
+
+Separate from the standard `alibaba` profile because it hits a different
+endpoint (coding-intl.dashscope.aliyuncs.com) with a dedicated API key tier.
+"""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+alibaba_coding_plan = ProviderProfile(
+    name="alibaba-coding-plan",
+    aliases=("alibaba_coding", "alibaba-coding", "dashscope-coding"),
+    display_name="Alibaba Cloud (Coding Plan)",
+    description="Alibaba Cloud Coding Plan — dedicated coding tier",
+    signup_url="https://help.aliyun.com/zh/model-studio/",
+    env_vars=("ALIBABA_CODING_PLAN_API_KEY", "DASHSCOPE_API_KEY", "ALIBABA_CODING_PLAN_BASE_URL"),
+    base_url="https://coding-intl.dashscope.aliyuncs.com/v1",
+    auth_type="api_key",
+)
+
+register_provider(alibaba_coding_plan)

plugins/model-providers/alibaba-coding-plan/plugin.yaml

@@ -0,0 +1,5 @@
+name: alibaba-coding-plan-provider
+kind: model-provider
+version: 1.0.0
+description: Alibaba Cloud Coding Plan
+author: Nous Research

plugins/model-providers/alibaba/__init__.py

@@ -0,0 +1,13 @@
+"""Alibaba Cloud DashScope provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+alibaba = ProviderProfile(
+    name="alibaba",
+    aliases=("dashscope", "alibaba-cloud", "qwen-dashscope"),
+    env_vars=("DASHSCOPE_API_KEY",),
+    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
+)
+
+register_provider(alibaba)

plugins/model-providers/alibaba/plugin.yaml

@@ -0,0 +1,5 @@
+name: alibaba-provider
+kind: model-provider
+version: 1.0.0
+description: Alibaba DashScope (international)
+author: Nous Research

plugins/model-providers/anthropic/__init__.py

@@ -0,0 +1,52 @@
+"""Native Anthropic provider profile."""
+
+import json
+import logging
+import urllib.request
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+logger = logging.getLogger(__name__)
+
+
+class AnthropicProfile(ProviderProfile):
+    """Native Anthropic — uses x-api-key header, not Bearer."""
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Anthropic uses x-api-key header and anthropic-version."""
+        if not api_key:
+            return None
+        try:
+            req = urllib.request.Request("https://api.anthropic.com/v1/models")
+            req.add_header("x-api-key", api_key)
+            req.add_header("anthropic-version", "2023-06-01")
+            req.add_header("Accept", "application/json")
+            with urllib.request.urlopen(req, timeout=timeout) as resp:
+                data = json.loads(resp.read().decode())
+            return [
+                m["id"]
+                for m in data.get("data", [])
+                if isinstance(m, dict) and "id" in m
+            ]
+        except Exception as exc:
+            logger.debug("fetch_models(anthropic): %s", exc)
+            return None
+
+
+anthropic = AnthropicProfile(
+    name="anthropic",
+    aliases=("claude", "claude-oauth", "claude-code"),
+    api_mode="anthropic_messages",
+    env_vars=("ANTHROPIC_API_KEY", "ANTHROPIC_TOKEN", "CLAUDE_CODE_OAUTH_TOKEN"),
+    base_url="https://api.anthropic.com",
+    auth_type="api_key",
+    default_aux_model="claude-haiku-4-5-20251001",
+)
+
+register_provider(anthropic)

plugins/model-providers/anthropic/plugin.yaml

@@ -0,0 +1,5 @@
+name: anthropic-provider
+kind: model-provider
+version: 1.0.0
+description: Anthropic (Claude)
+author: Nous Research

plugins/model-providers/arcee/__init__.py

@@ -0,0 +1,13 @@
+"""Arcee AI provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+arcee = ProviderProfile(
+    name="arcee",
+    aliases=("arcee-ai", "arceeai"),
+    env_vars=("ARCEEAI_API_KEY",),
+    base_url="https://api.arcee.ai/api/v1",
+)
+
+register_provider(arcee)

plugins/model-providers/arcee/plugin.yaml

@@ -0,0 +1,5 @@
+name: arcee-provider
+kind: model-provider
+version: 1.0.0
+description: Arcee AI
+author: Nous Research

plugins/model-providers/azure-foundry/__init__.py

@@ -0,0 +1,21 @@
+"""Azure AI Foundry provider profile.
+
+Azure Foundry exposes an OpenAI-compatible endpoint; users supply their own
+base URL at setup since endpoints are per-resource.
+"""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+azure_foundry = ProviderProfile(
+    name="azure-foundry",
+    aliases=("azure", "azure-ai-foundry", "azure-ai"),
+    display_name="Azure Foundry",
+    description="Azure AI Foundry — OpenAI-compatible endpoint (user-supplied base URL)",
+    signup_url="https://ai.azure.com/",
+    env_vars=("AZURE_FOUNDRY_API_KEY", "AZURE_FOUNDRY_BASE_URL"),
+    base_url="",  # per-resource; user provides at setup
+    auth_type="api_key",
+)
+
+register_provider(azure_foundry)

plugins/model-providers/azure-foundry/plugin.yaml

@@ -0,0 +1,5 @@
+name: azure-foundry-provider
+kind: model-provider
+version: 1.0.0
+description: Azure AI Foundry
+author: Nous Research

plugins/model-providers/bedrock/__init__.py

@@ -0,0 +1,29 @@
+"""AWS Bedrock provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class BedrockProfile(ProviderProfile):
+    """AWS Bedrock — no REST /v1/models endpoint; uses AWS SDK."""
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Bedrock model listing requires AWS SDK, not a REST call."""
+        return None
+
+
+bedrock = BedrockProfile(
+    name="bedrock",
+    aliases=("aws", "aws-bedrock", "amazon-bedrock", "amazon"),
+    api_mode="bedrock_converse",
+    env_vars=(),  # AWS SDK credentials — not env vars
+    base_url="https://bedrock-runtime.us-east-1.amazonaws.com",
+    auth_type="aws_sdk",
+)
+
+register_provider(bedrock)

plugins/model-providers/bedrock/plugin.yaml

@@ -0,0 +1,5 @@
+name: bedrock-provider
+kind: model-provider
+version: 1.0.0
+description: AWS Bedrock
+author: Nous Research

plugins/model-providers/copilot-acp/__init__.py

@@ -0,0 +1,34 @@
+"""GitHub Copilot ACP provider profile.
+
+copilot-acp uses an external ACP subprocess — NOT the standard
+transport. api_mode="copilot_acp" is handled separately in run_agent.py.
+The profile captures auth + endpoint metadata for registry migration.
+"""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class CopilotACPProfile(ProviderProfile):
+    """GitHub Copilot ACP — external process, no REST models endpoint."""
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Model listing is handled by the ACP subprocess."""
+        return None
+
+
+copilot_acp = CopilotACPProfile(
+    name="copilot-acp",
+    aliases=("github-copilot-acp", "copilot-acp-agent"),
+    api_mode="chat_completions",  # ACP subprocess uses chat_completions routing
+    env_vars=(),  # Managed by ACP subprocess
+    base_url="acp://copilot",  # ACP internal scheme
+    auth_type="external_process",
+)
+
+register_provider(copilot_acp)

plugins/model-providers/copilot-acp/plugin.yaml

@@ -0,0 +1,5 @@
+name: copilot-acp-provider
+kind: model-provider
+version: 1.0.0
+description: GitHub Copilot via ACP subprocess
+author: Nous Research

plugins/model-providers/copilot/__init__.py

@@ -0,0 +1,58 @@
+"""Copilot / GitHub Models provider profile.
+
+Copilot uses per-model api_mode routing:
+  - GPT-5+ / Codex models → codex_responses
+  - Claude models → anthropic_messages
+  - Everything else → chat_completions (this profile covers that subset)
+
+Key quirks for the chat_completions subset:
+  - Editor attribution headers (via copilot_default_headers())
+  - GitHub Models reasoning extra_body (model-catalog gated)
+"""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class CopilotProfile(ProviderProfile):
+    """GitHub Copilot / GitHub Models — editor headers + reasoning."""
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        model: str | None = None,
+        reasoning_config: dict | None = None,
+        supports_reasoning: bool = False,
+        **ctx,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        extra_body: dict[str, Any] = {}
+        if supports_reasoning and model:
+            try:
+                from hermes_cli.models import github_model_reasoning_efforts
+
+                supported_efforts = github_model_reasoning_efforts(model)
+                if supported_efforts and reasoning_config:
+                    effort = reasoning_config.get("effort", "medium")
+                    # Normalize non-standard effort levels to the nearest supported
+                    if effort == "xhigh":
+                        effort = "high"
+                    if effort in supported_efforts:
+                        extra_body["reasoning"] = {"effort": effort}
+                elif supported_efforts:
+                    extra_body["reasoning"] = {"effort": "medium"}
+            except Exception:
+                pass
+        return extra_body, {}
+
+
+copilot = CopilotProfile(
+    name="copilot",
+    aliases=("github-copilot", "github-models", "github-model", "github"),
+    env_vars=("COPILOT_GITHUB_TOKEN", "GH_TOKEN", "GITHUB_TOKEN"),
+    base_url="https://api.githubcopilot.com",
+    auth_type="copilot",
+)
+
+register_provider(copilot)

plugins/model-providers/copilot/plugin.yaml

@@ -0,0 +1,5 @@
+name: copilot-provider
+kind: model-provider
+version: 1.0.0
+description: GitHub Copilot
+author: Nous Research

plugins/model-providers/custom/__init__.py

@@ -0,0 +1,68 @@
+"""Custom / Ollama (local) provider profile.
+
+Covers any endpoint registered as provider="custom", including local
+Ollama instances. Key quirks:
+  - ollama_num_ctx → extra_body.options.num_ctx (local context window)
+  - reasoning_config disabled → extra_body.think = False
+"""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class CustomProfile(ProviderProfile):
+    """Custom/Ollama local provider — think=false and num_ctx support."""
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        ollama_num_ctx: int | None = None,
+        **ctx: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        extra_body: dict[str, Any] = {}
+
+        # Ollama context window
+        if ollama_num_ctx:
+            options = extra_body.get("options", {})
+            options["num_ctx"] = ollama_num_ctx
+            extra_body["options"] = options
+
+        # Disable thinking when reasoning is turned off
+        if reasoning_config and isinstance(reasoning_config, dict):
+            _effort = (reasoning_config.get("effort") or "").strip().lower()
+            _enabled = reasoning_config.get("enabled", True)
+            if _effort == "none" or _enabled is False:
+                extra_body["think"] = False
+
+        return extra_body, {}
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Custom/Ollama: base_url is user-configured; fetch if set."""
+        if not self.base_url:
+            return None
+        return super().fetch_models(api_key=api_key, timeout=timeout)
+
+
+custom = CustomProfile(
+    name="custom",
+    aliases=(
+        "ollama",
+        "local",
+        "vllm",
+        "llamacpp",
+        "llama.cpp",
+        "llama-cpp",
+    ),
+    env_vars=(),  # No fixed key — custom endpoint
+    base_url="",  # User-configured
+)
+
+register_provider(custom)

plugins/model-providers/custom/plugin.yaml

@@ -0,0 +1,5 @@
+name: custom-provider
+kind: model-provider
+version: 1.0.0
+description: Custom / Ollama / local OpenAI-compatible endpoint
+author: Nous Research

plugins/model-providers/deepseek/__init__.py

@@ -0,0 +1,20 @@
+"""DeepSeek provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+deepseek = ProviderProfile(
+    name="deepseek",
+    aliases=("deepseek-chat",),
+    env_vars=("DEEPSEEK_API_KEY",),
+    display_name="DeepSeek",
+    description="DeepSeek — native DeepSeek API",
+    signup_url="https://platform.deepseek.com/",
+    fallback_models=(
+        "deepseek-chat",
+        "deepseek-reasoner",
+    ),
+    base_url="https://api.deepseek.com/v1",
+)
+
+register_provider(deepseek)

plugins/model-providers/deepseek/plugin.yaml

@@ -0,0 +1,5 @@
+name: deepseek-provider
+kind: model-provider
+version: 1.0.0
+description: DeepSeek
+author: Nous Research

plugins/model-providers/gemini/__init__.py

@@ -0,0 +1,72 @@
+"""Google Gemini provider profiles.
+
+gemini:            Google AI Studio (API key) — uses GeminiNativeClient
+google-gemini-cli: Google Cloud Code Assist (OAuth) — uses GeminiCloudCodeClient
+
+Both report api_mode="chat_completions" but use custom native clients
+that bypass the standard OpenAI transport. The profile captures auth
+and endpoint metadata for auth.py / runtime_provider.py migration, and
+carries the thinking_config translation hook so the transport's profile
+path produces the same extra_body shape the legacy flag path did.
+"""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class GeminiProfile(ProviderProfile):
+    """Gemini — translate reasoning_config to thinking_config in extra_body."""
+
+    def build_extra_body(
+        self, *, session_id: str | None = None, **context: Any
+    ) -> dict[str, Any]:
+        """Emit extra_body.thinking_config (native) or extra_body.extra_body.google.thinking_config
+        (OpenAI-compat /openai subpath), mirroring the legacy path's behavior.
+        """
+        from agent.transports.chat_completions import (
+            _build_gemini_thinking_config,
+            _is_gemini_openai_compat_base_url,
+            _snake_case_gemini_thinking_config,
+        )
+
+        model = context.get("model") or ""
+        reasoning_config = context.get("reasoning_config")
+        base_url = context.get("base_url") or self.base_url
+
+        raw_thinking_config = _build_gemini_thinking_config(model, reasoning_config)
+        if not raw_thinking_config:
+            return {}
+
+        body: dict[str, Any] = {}
+        if self.name == "gemini" and _is_gemini_openai_compat_base_url(base_url):
+            thinking_config = _snake_case_gemini_thinking_config(raw_thinking_config)
+            if thinking_config:
+                body["extra_body"] = {"google": {"thinking_config": thinking_config}}
+        else:
+            body["thinking_config"] = raw_thinking_config
+        return body
+
+
+gemini = GeminiProfile(
+    name="gemini",
+    aliases=("google", "google-gemini", "google-ai-studio"),
+    api_mode="chat_completions",
+    env_vars=("GOOGLE_API_KEY", "GEMINI_API_KEY"),
+    base_url="https://generativelanguage.googleapis.com/v1beta",
+    auth_type="api_key",
+    default_aux_model="gemini-3-flash-preview",
+)
+
+google_gemini_cli = GeminiProfile(
+    name="google-gemini-cli",
+    aliases=("gemini-cli", "gemini-oauth"),
+    api_mode="chat_completions",
+    env_vars=(),  # OAuth — no API key
+    base_url="cloudcode-pa://google",  # Cloud Code Assist internal scheme
+    auth_type="oauth_external",
+)
+
+register_provider(gemini)
+register_provider(google_gemini_cli)

plugins/model-providers/gemini/plugin.yaml

@@ -0,0 +1,5 @@
+name: gemini-provider
+kind: model-provider
+version: 1.0.0
+description: Google Gemini (API key + Cloud Code OAuth)
+author: Nous Research

plugins/model-providers/gmi/__init__.py

@@ -0,0 +1,26 @@
+"""GMI Cloud provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+gmi = ProviderProfile(
+    name="gmi",
+    aliases=("gmi-cloud", "gmicloud"),
+    display_name="GMI Cloud",
+    description="GMI Cloud — multi-model direct API (slash-form model IDs)",
+    signup_url="https://www.gmicloud.ai/",
+    env_vars=("GMI_API_KEY", "GMI_BASE_URL"),
+    base_url="https://api.gmi-serving.com/v1",
+    auth_type="api_key",
+    default_aux_model="google/gemini-3.1-flash-lite-preview",
+    fallback_models=(
+        "zai-org/GLM-5.1-FP8",
+        "deepseek-ai/DeepSeek-V3.2",
+        "moonshotai/Kimi-K2.5",
+        "google/gemini-3.1-flash-lite-preview",
+        "anthropic/claude-sonnet-4.6",
+        "openai/gpt-5.4",
+    ),
+)
+
+register_provider(gmi)

plugins/model-providers/gmi/plugin.yaml

@@ -0,0 +1,5 @@
+name: gmi-provider
+kind: model-provider
+version: 1.0.0
+description: GMI Cloud
+author: Nous Research

plugins/model-providers/huggingface/__init__.py

@@ -0,0 +1,20 @@
+"""Hugging Face provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+huggingface = ProviderProfile(
+    name="huggingface",
+    aliases=("hf", "hugging-face", "huggingface-hub"),
+    env_vars=("HF_TOKEN",),
+    display_name="HuggingFace",
+    description="HuggingFace Inference API",
+    signup_url="https://huggingface.co/settings/tokens",
+    fallback_models=(
+        "Qwen/Qwen3.5-72B-Instruct",
+        "deepseek-ai/DeepSeek-V3.2",
+    ),
+    base_url="https://router.huggingface.co/v1",
+)
+
+register_provider(huggingface)

plugins/model-providers/huggingface/plugin.yaml

@@ -0,0 +1,5 @@
+name: huggingface-provider
+kind: model-provider
+version: 1.0.0
+description: HuggingFace Inference Providers
+author: Nous Research

plugins/model-providers/kilocode/__init__.py

@@ -0,0 +1,14 @@
+"""Kilo Code provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+kilocode = ProviderProfile(
+    name="kilocode",
+    aliases=("kilo-code", "kilo", "kilo-gateway"),
+    env_vars=("KILOCODE_API_KEY",),
+    base_url="https://api.kilo.ai/api/gateway",
+    default_aux_model="google/gemini-3-flash-preview",
+)
+
+register_provider(kilocode)

plugins/model-providers/kilocode/plugin.yaml

@@ -0,0 +1,5 @@
+name: kilocode-provider
+kind: model-provider
+version: 1.0.0
+description: Kilo Code
+author: Nous Research

plugins/model-providers/kimi-coding/__init__.py

@@ -0,0 +1,71 @@
+"""Kimi / Moonshot provider profiles.
+
+Kimi has dual endpoints:
+  - sk-kimi-* keys → api.kimi.com/coding (Anthropic Messages API)
+  - legacy keys → api.moonshot.ai/v1 (OpenAI chat completions)
+
+This module covers the chat_completions path (/v1 endpoint).
+"""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import OMIT_TEMPERATURE, ProviderProfile
+
+
+class KimiProfile(ProviderProfile):
+    """Kimi/Moonshot — temperature omitted, thinking + reasoning_effort."""
+
+    def build_api_kwargs_extras(
+        self, *, reasoning_config: dict | None = None, **context
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """Kimi uses extra_body.thinking + top-level reasoning_effort."""
+        extra_body = {}
+        top_level = {}
+
+        if not reasoning_config or not isinstance(reasoning_config, dict):
+            # No config → thinking enabled, default effort
+            extra_body["thinking"] = {"type": "enabled"}
+            top_level["reasoning_effort"] = "medium"
+            return extra_body, top_level
+
+        enabled = reasoning_config.get("enabled", True)
+        if enabled is False:
+            extra_body["thinking"] = {"type": "disabled"}
+            return extra_body, top_level
+
+        # Enabled
+        extra_body["thinking"] = {"type": "enabled"}
+        effort = (reasoning_config.get("effort") or "").strip().lower()
+        if effort in ("low", "medium", "high"):
+            top_level["reasoning_effort"] = effort
+        else:
+            top_level["reasoning_effort"] = "medium"
+
+        return extra_body, top_level
+
+
+kimi = KimiProfile(
+    name="kimi-coding",
+    aliases=("kimi", "moonshot", "kimi-for-coding"),
+    env_vars=("KIMI_API_KEY", "KIMI_CODING_API_KEY"),
+    base_url="https://api.moonshot.ai/v1",
+    fixed_temperature=OMIT_TEMPERATURE,
+    default_max_tokens=32000,
+    default_headers={"User-Agent": "hermes-agent/1.0"},
+    default_aux_model="kimi-k2-turbo-preview",
+)
+
+kimi_cn = KimiProfile(
+    name="kimi-coding-cn",
+    aliases=("kimi-cn", "moonshot-cn"),
+    env_vars=("KIMI_CN_API_KEY",),
+    base_url="https://api.moonshot.cn/v1",
+    fixed_temperature=OMIT_TEMPERATURE,
+    default_max_tokens=32000,
+    default_headers={"User-Agent": "hermes-agent/1.0"},
+    default_aux_model="kimi-k2-turbo-preview",
+)
+
+register_provider(kimi)
+register_provider(kimi_cn)

plugins/model-providers/kimi-coding/plugin.yaml

@@ -0,0 +1,5 @@
+name: kimi-coding-provider
+kind: model-provider
+version: 1.0.0
+description: Moonshot Kimi Coding (global + China)
+author: Nous Research

plugins/model-providers/minimax/__init__.py

@@ -0,0 +1,45 @@
+"""MiniMax provider profiles (international + China).
+
+Both use anthropic_messages api_mode — their inference_base_url
+ends with /anthropic which triggers auto-detection to anthropic_messages.
+"""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+minimax = ProviderProfile(
+    name="minimax",
+    aliases=("mini-max",),
+    api_mode="anthropic_messages",
+    env_vars=("MINIMAX_API_KEY",),
+    base_url="https://api.minimax.io/anthropic",
+    auth_type="api_key",
+    default_aux_model="MiniMax-M2.7",
+)
+
+minimax_cn = ProviderProfile(
+    name="minimax-cn",
+    aliases=("minimax-china", "minimax_cn"),
+    api_mode="anthropic_messages",
+    env_vars=("MINIMAX_CN_API_KEY",),
+    base_url="https://api.minimaxi.com/anthropic",
+    auth_type="api_key",
+    default_aux_model="MiniMax-M2.7",
+)
+
+minimax_oauth = ProviderProfile(
+    name="minimax-oauth",
+    aliases=("minimax_oauth", "minimax-oauth-io"),
+    api_mode="anthropic_messages",
+    display_name="MiniMax (OAuth)",
+    description="MiniMax via OAuth browser flow — no API key required",
+    signup_url="https://api.minimax.io/",
+    env_vars=(),  # OAuth — tokens in auth.json, not env
+    base_url="https://api.minimax.io/anthropic",
+    auth_type="oauth_external",
+    default_aux_model="MiniMax-M2.7-highspeed",
+)
+
+register_provider(minimax)
+register_provider(minimax_cn)
+register_provider(minimax_oauth)

plugins/model-providers/minimax/plugin.yaml

@@ -0,0 +1,5 @@
+name: minimax-provider
+kind: model-provider
+version: 1.0.0
+description: MiniMax M-series (global + China + OAuth)
+author: Nous Research

plugins/model-providers/nous/__init__.py

@@ -0,0 +1,53 @@
+"""Nous Portal provider profile."""
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class NousProfile(ProviderProfile):
+    """Nous Portal — product tags, reasoning with Nous-specific omission."""
+
+    def build_extra_body(
+        self, *, session_id: str | None = None, **context
+    ) -> dict[str, Any]:
+        return {"tags": ["product=hermes-agent"]}
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        supports_reasoning: bool = False,
+        **context,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """Nous: passes full reasoning_config, but OMITS when disabled."""
+        extra_body = {}
+        if supports_reasoning:
+            if reasoning_config is not None:
+                rc = dict(reasoning_config)
+                if rc.get("enabled") is False:
+                    pass  # Nous omits reasoning when disabled
+                else:
+                    extra_body["reasoning"] = rc
+            else:
+                extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
+        return extra_body, {}
+
+
+nous = NousProfile(
+    name="nous",
+    aliases=("nous-portal", "nousresearch"),
+    env_vars=("NOUS_API_KEY",),
+    display_name="Nous Research",
+    description="Nous Research — Hermes model family",
+    signup_url="https://nousresearch.com/",
+    fallback_models=(
+        "hermes-3-405b",
+        "hermes-3-70b",
+    ),
+    base_url="https://inference.nousresearch.com/v1",
+    auth_type="oauth_device_code",
+)
+
+register_provider(nous)

plugins/model-providers/nous/plugin.yaml

@@ -0,0 +1,5 @@
+name: nous-provider
+kind: model-provider
+version: 1.0.0
+description: Nous Research Portal
+author: Nous Research

plugins/model-providers/nvidia/__init__.py

@@ -0,0 +1,21 @@
+"""NVIDIA NIM provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+nvidia = ProviderProfile(
+    name="nvidia",
+    aliases=("nvidia-nim",),
+    env_vars=("NVIDIA_API_KEY",),
+    display_name="NVIDIA NIM",
+    description="NVIDIA NIM — accelerated inference",
+    signup_url="https://build.nvidia.com/",
+    fallback_models=(
+        "nvidia/llama-3.1-nemotron-70b-instruct",
+        "nvidia/llama-3.3-70b-instruct",
+    ),
+    base_url="https://integrate.api.nvidia.com/v1",
+    default_max_tokens=16384,
+)
+
+register_provider(nvidia)

plugins/model-providers/nvidia/plugin.yaml

@@ -0,0 +1,5 @@
+name: nvidia-provider
+kind: model-provider
+version: 1.0.0
+description: NVIDIA NIM
+author: Nous Research

plugins/model-providers/ollama-cloud/__init__.py

@@ -0,0 +1,14 @@
+"""Ollama Cloud provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+ollama_cloud = ProviderProfile(
+    name="ollama-cloud",
+    aliases=("ollama_cloud",),
+    default_aux_model="nemotron-3-nano:30b",
+    env_vars=("OLLAMA_API_KEY",),
+    base_url="https://ollama.com/v1",
+)
+
+register_provider(ollama_cloud)

plugins/model-providers/ollama-cloud/plugin.yaml

@@ -0,0 +1,5 @@
+name: ollama-cloud-provider
+kind: model-provider
+version: 1.0.0
+description: Ollama Cloud
+author: Nous Research

plugins/model-providers/openai-codex/__init__.py

@@ -0,0 +1,15 @@
+"""OpenAI Codex (Responses API) provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+openai_codex = ProviderProfile(
+    name="openai-codex",
+    aliases=("codex", "openai_codex"),
+    api_mode="codex_responses",
+    env_vars=(),  # OAuth external — no API key
+    base_url="https://chatgpt.com/backend-api/codex",
+    auth_type="oauth_external",
+)
+
+register_provider(openai_codex)

plugins/model-providers/openai-codex/plugin.yaml

@@ -0,0 +1,5 @@
+name: openai-codex-provider
+kind: model-provider
+version: 1.0.0
+description: OpenAI Codex (Responses API)
+author: Nous Research

plugins/model-providers/opencode-zen/__init__.py

@@ -0,0 +1,30 @@
+"""OpenCode provider profiles (Zen + Go).
+
+Both use per-model api_mode routing:
+  - OpenCode Zen: Claude → anthropic_messages, GPT-5/Codex → codex_responses,
+    everything else → chat_completions (this profile)
+  - OpenCode Go: MiniMax → anthropic_messages, GLM/Kimi → chat_completions
+    (this profile)
+"""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+opencode_zen = ProviderProfile(
+    name="opencode-zen",
+    aliases=("opencode", "opencode_zen", "zen"),
+    env_vars=("OPENCODE_ZEN_API_KEY",),
+    base_url="https://opencode.ai/zen/v1",
+    default_aux_model="gemini-3-flash",
+)
+
+opencode_go = ProviderProfile(
+    name="opencode-go",
+    aliases=("opencode_go", "go", "opencode-go-sub"),
+    env_vars=("OPENCODE_GO_API_KEY",),
+    base_url="https://opencode.ai/zen/go/v1",
+    default_aux_model="glm-5",
+)
+
+register_provider(opencode_zen)
+register_provider(opencode_go)

plugins/model-providers/opencode-zen/plugin.yaml

@@ -0,0 +1,5 @@
+name: opencode-zen-provider
+kind: model-provider
+version: 1.0.0
+description: OpenCode (Zen + Go)
+author: Nous Research

plugins/model-providers/openrouter/__init__.py

@@ -0,0 +1,86 @@
+"""OpenRouter provider profile."""
+
+import logging
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+logger = logging.getLogger(__name__)
+
+_CACHE: list[str] | None = None
+
+
+class OpenRouterProfile(ProviderProfile):
+    """OpenRouter aggregator — provider preferences, reasoning config passthrough."""
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Fetch from public OpenRouter catalog — no auth required.
+
+        Note: Tool-call capability filtering is applied by hermes_cli/models.py
+        via fetch_openrouter_models() → _openrouter_model_supports_tools(), not
+        here. The picker early-returns via the dedicated openrouter path before
+        reaching this method, so filtering here would be unreachable.
+        """
+        global _CACHE  # noqa: PLW0603
+        if _CACHE is not None:
+            return _CACHE
+        try:
+            result = super().fetch_models(api_key=None, timeout=timeout)
+            if result is not None:
+                _CACHE = result
+            return result
+        except Exception as exc:
+            logger.debug("fetch_models(openrouter): %s", exc)
+            return None
+
+    def build_extra_body(
+        self, *, session_id: str | None = None, **context: Any
+    ) -> dict[str, Any]:
+        body: dict[str, Any] = {}
+        prefs = context.get("provider_preferences")
+        if prefs:
+            body["provider"] = prefs
+        return body
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        supports_reasoning: bool = False,
+        **context: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """OpenRouter passes the full reasoning_config dict as extra_body.reasoning."""
+        extra_body: dict[str, Any] = {}
+        if supports_reasoning:
+            if reasoning_config is not None:
+                extra_body["reasoning"] = dict(reasoning_config)
+            else:
+                extra_body["reasoning"] = {"enabled": True, "effort": "medium"}
+        return extra_body, {}
+
+
+openrouter = OpenRouterProfile(
+    name="openrouter",
+    aliases=("or",),
+    env_vars=("OPENROUTER_API_KEY",),
+    display_name="OpenRouter",
+    description="OpenRouter — unified API for 200+ models",
+    signup_url="https://openrouter.ai/keys",
+    base_url="https://openrouter.ai/api/v1",
+    models_url="https://openrouter.ai/api/v1/models",
+    fallback_models=(
+        "anthropic/claude-sonnet-4.6",
+        "openai/gpt-5.4",
+        "deepseek/deepseek-chat",
+        "google/gemini-3-flash-preview",
+        "qwen/qwen3-plus",
+    ),
+)
+
+register_provider(openrouter)

plugins/model-providers/openrouter/plugin.yaml

@@ -0,0 +1,5 @@
+name: openrouter-provider
+kind: model-provider
+version: 1.0.0
+description: OpenRouter aggregator
+author: Nous Research

plugins/model-providers/qwen-oauth/__init__.py

@@ -0,0 +1,82 @@
+"""Qwen Portal provider profile."""
+
+import copy
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+class QwenProfile(ProviderProfile):
+    """Qwen Portal — message normalization, vl_high_resolution, metadata top-level."""
+
+    def prepare_messages(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Normalize content to list-of-dicts format.
+
+        Inject cache_control on system message.
+
+        Matches the behavior of run_agent.py:_qwen_prepare_chat_messages().
+        """
+        prepared = copy.deepcopy(messages)
+        if not prepared:
+            return prepared
+
+        for msg in prepared:
+            if not isinstance(msg, dict):
+                continue
+            content = msg.get("content")
+            if isinstance(content, str):
+                msg["content"] = [{"type": "text", "text": content}]
+            elif isinstance(content, list):
+                normalized_parts = []
+                for part in content:
+                    if isinstance(part, str):
+                        normalized_parts.append({"type": "text", "text": part})
+                    elif isinstance(part, dict):
+                        normalized_parts.append(part)
+                if normalized_parts:
+                    msg["content"] = normalized_parts
+
+        # Inject cache_control on the last part of the system message.
+        for msg in prepared:
+            if isinstance(msg, dict) and msg.get("role") == "system":
+                content = msg.get("content")
+                if (
+                    isinstance(content, list)
+                    and content
+                    and isinstance(content[-1], dict)
+                ):
+                    content[-1]["cache_control"] = {"type": "ephemeral"}
+                break
+
+        return prepared
+
+    def build_extra_body(
+        self, *, session_id: str | None = None, **context
+    ) -> dict[str, Any]:
+        return {"vl_high_resolution_images": True}
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        qwen_session_metadata: dict | None = None,
+        **context,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """Qwen metadata goes to top-level api_kwargs, not extra_body."""
+        top_level = {}
+        if qwen_session_metadata:
+            top_level["metadata"] = qwen_session_metadata
+        return {}, top_level
+
+
+qwen = QwenProfile(
+    name="qwen-oauth",
+    aliases=("qwen", "qwen-portal", "qwen-cli"),
+    env_vars=("QWEN_API_KEY",),
+    base_url="https://portal.qwen.ai/v1",
+    auth_type="oauth_external",
+    default_max_tokens=65536,
+)
+
+register_provider(qwen)

plugins/model-providers/qwen-oauth/plugin.yaml

@@ -0,0 +1,5 @@
+name: qwen-oauth-provider
+kind: model-provider
+version: 1.0.0
+description: Qwen Portal (OAuth)
+author: Nous Research

plugins/model-providers/stepfun/__init__.py

@@ -0,0 +1,14 @@
+"""StepFun provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+stepfun = ProviderProfile(
+    name="stepfun",
+    aliases=("step", "stepfun-coding-plan"),
+    default_aux_model="step-3.5-flash",
+    env_vars=("STEPFUN_API_KEY",),
+    base_url="https://api.stepfun.ai/step_plan/v1",
+)
+
+register_provider(stepfun)

plugins/model-providers/stepfun/plugin.yaml

@@ -0,0 +1,5 @@
+name: stepfun-provider
+kind: model-provider
+version: 1.0.0
+description: StepFun Step Plan
+author: Nous Research

plugins/model-providers/xai/__init__.py

@@ -0,0 +1,15 @@
+"""xAI (Grok) provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+xai = ProviderProfile(
+    name="xai",
+    aliases=("grok", "x-ai", "x.ai"),
+    api_mode="codex_responses",
+    env_vars=("XAI_API_KEY",),
+    base_url="https://api.x.ai/v1",
+    auth_type="api_key",
+)
+
+register_provider(xai)

plugins/model-providers/xai/plugin.yaml

@@ -0,0 +1,5 @@
+name: xai-provider
+kind: model-provider
+version: 1.0.0
+description: xAI Grok (Responses API)
+author: Nous Research

plugins/model-providers/xiaomi/__init__.py

@@ -0,0 +1,13 @@
+"""Xiaomi MiMo provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+xiaomi = ProviderProfile(
+    name="xiaomi",
+    aliases=("mimo", "xiaomi-mimo"),
+    env_vars=("XIAOMI_API_KEY",),
+    base_url="https://api.xiaomimimo.com/v1",
+)
+
+register_provider(xiaomi)

plugins/model-providers/xiaomi/plugin.yaml

@@ -0,0 +1,5 @@
+name: xiaomi-provider
+kind: model-provider
+version: 1.0.0
+description: Xiaomi MiMo
+author: Nous Research

plugins/model-providers/zai/__init__.py

@@ -0,0 +1,21 @@
+"""ZAI / GLM provider profile."""
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+zai = ProviderProfile(
+    name="zai",
+    aliases=("glm", "z-ai", "z.ai", "zhipu"),
+    env_vars=("GLM_API_KEY", "ZAI_API_KEY", "Z_AI_API_KEY"),
+    display_name="Z.AI (GLM)",
+    description="Z.AI / GLM — Zhipu AI models",
+    signup_url="https://z.ai/",
+    fallback_models=(
+        "glm-5",
+        "glm-4-9b",
+    ),
+    base_url="https://api.z.ai/api/paas/v4",
+    default_aux_model="glm-4.5-flash",
+)
+
+register_provider(zai)

plugins/model-providers/zai/plugin.yaml

@@ -0,0 +1,5 @@
+name: zai-provider
+kind: model-provider
+version: 1.0.0
+description: Z.AI / GLM
+author: Nous Research

providers/README.md

@@ -0,0 +1,78 @@
+# providers/
+
+Registry and ABC for every inference provider Hermes knows about.
+
+Each provider is declared once as a `ProviderProfile`. Every other layer —
+auth resolution, transport kwargs, model listing, runtime routing — reads from
+these profiles instead of maintaining its own parallel data.
+
+---
+
+## Layout
+
+```
+providers/
+├── base.py         ProviderProfile dataclass + OMIT_TEMPERATURE sentinel
+├── __init__.py     Registry: register_provider(), get_provider_profile(), list_providers()
+└── README.md       This file
+```
+
+The **profiles themselves** live as plugins under
+`plugins/model-providers/<name>/` (bundled in this repo) and
+`$HERMES_HOME/plugins/model-providers/<name>/` (per-user overrides). The
+registry in `providers/__init__.py` lazily discovers them the first time any
+consumer calls `get_provider_profile()` or `list_providers()`. See
+`plugins/model-providers/README.md` for the plugin contract and examples.
+
+---
+
+## How it wires in
+
+The registry is populated on first access. After that, every downstream
+layer reads from it:
+
+- `hermes_cli/auth.py` extends `PROVIDER_REGISTRY` with every api-key
+  profile it sees (skipping `copilot`, `kimi-coding`, `kimi-coding-cn`,
+  `zai`, `openrouter`, `custom` — those need bespoke token resolution).
+- `hermes_cli/models.py` extends `CANONICAL_PROVIDERS` and calls
+  `profile.fetch_models()` inside `provider_model_ids()`.
+- `hermes_cli/doctor.py` adds a `/models` health check for each
+  `auth_type="api_key"` profile.
+- `hermes_cli/config.py` injects every `env_var` into
+  `OPTIONAL_ENV_VARS` so the setup wizard knows about it.
+- `hermes_cli/runtime_provider.py` reads `profile.api_mode` as a fallback
+  when URL detection finds nothing.
+- `agent/model_metadata.py` maps hostname → provider via
+  `profile.get_hostname()`.
+- `agent/auxiliary_client.py` reads `profile.default_aux_model` first
+  before falling back to the legacy hardcoded dict.
+- `agent/transports/chat_completions.py::_build_kwargs_from_profile()`
+  invokes `profile.prepare_messages()`, `profile.build_extra_body()`,
+  and `profile.build_api_kwargs_extras()` on every call.
+- `run_agent.py` passes `provider_profile=<ProviderProfile>` so the
+  transport takes the profile path instead of the legacy flag path.
+
+---
+
+## Adding a provider
+
+See `plugins/model-providers/README.md` — drop a new directory there (or
+under `$HERMES_HOME/plugins/model-providers/` for a private plugin).
+
+---
+
+## Hooks you can override on `ProviderProfile`
+
+| Hook | Purpose |
+|------|---------|
+| `get_hostname()` | URL-based detection — default derives from `base_url`. |
+| `prepare_messages(msgs)` | Provider-specific message preprocessing (Qwen normalises to list-of-parts, injects `cache_control`). |
+| `build_extra_body(**ctx)` | Provider-specific `extra_body` (OpenRouter provider prefs, Gemini `thinking_config`). |
+| `build_api_kwargs_extras(**ctx)` | `(extra_body_additions, top_level_kwargs)` — Kimi puts reasoning_effort top-level, Qwen splits `enable_thinking`/`thinking_budget`. |
+| `fetch_models(*, api_key)` | Live catalog fetch — default hits `{models_url or base_url}/models` with Bearer auth. Override for no-REST providers (Bedrock), OAuth catalogs (Anthropic), or public catalogs (OpenRouter). |
+
+---
+
+## Configuration fields
+
+Full reference in `providers/base.py` dataclass definition.

providers/__init__.py

@@ -0,0 +1,191 @@
+"""Provider module registry.
+
+Provider profiles can live in two places:
+
+1. Bundled plugins: ``plugins/model-providers/<name>/`` (shipped with hermes-agent)
+2. User plugins: ``$HERMES_HOME/plugins/model-providers/<name>/``
+
+Each plugin directory contains:
+  - ``__init__.py`` — calls ``register_provider(profile)`` at import
+  - ``plugin.yaml`` — manifest (name, kind: model-provider, version, description)
+
+Discovery is lazy: the first call to ``get_provider_profile()`` or
+``list_providers()`` scans both locations and imports every plugin. User
+plugins override bundled plugins on name collision (last-writer-wins), so
+third parties can monkey-patch or replace any built-in profile without
+editing the repo.
+
+For backward compatibility, ``providers/*.py`` files (other than ``base.py``
+and ``__init__.py``) are still discovered via ``pkgutil.iter_modules``.
+This lets out-of-tree users drop a single-file profile into an editable
+install without the plugin dir structure. New profiles should prefer the
+plugin layout.
+
+Usage::
+
+    from providers import get_provider_profile
+    profile = get_provider_profile("nvidia")   # ProviderProfile or None
+    profile = get_provider_profile("kimi")     # checks name + aliases
+"""
+
+from __future__ import annotations
+
+import importlib
+import importlib.util
+import logging
+import sys
+from pathlib import Path
+
+from providers.base import OMIT_TEMPERATURE, ProviderProfile  # noqa: F401
+
+logger = logging.getLogger(__name__)
+
+_REGISTRY: dict[str, ProviderProfile] = {}
+_ALIASES: dict[str, str] = {}
+_discovered = False
+
+# Repo-root ``plugins/model-providers/`` — populated at discovery time.
+_BUNDLED_PLUGINS_DIR = (
+    Path(__file__).resolve().parent.parent / "plugins" / "model-providers"
+)
+
+
+def register_provider(profile: ProviderProfile) -> None:
+    """Register a provider profile by name and aliases.
+
+    Later registrations with the same name replace earlier ones — so user
+    plugins under ``$HERMES_HOME/plugins/model-providers/`` can override
+    bundled profiles without editing repo code.
+    """
+    _REGISTRY[profile.name] = profile
+    for alias in profile.aliases:
+        _ALIASES[alias] = profile.name
+
+
+def get_provider_profile(name: str) -> ProviderProfile | None:
+    """Look up a provider profile by name or alias.
+
+    Returns None if the provider has no profile (falls back to generic).
+    """
+    if not _discovered:
+        _discover_providers()
+    canonical = _ALIASES.get(name, name)
+    return _REGISTRY.get(canonical)
+
+
+def list_providers() -> list[ProviderProfile]:
+    """Return all registered provider profiles (one per canonical name)."""
+    if not _discovered:
+        _discover_providers()
+    # Deduplicate: _REGISTRY has canonical names; _ALIASES points to same objects
+    seen: set[int] = set()
+    result: list[ProviderProfile] = []
+    for profile in _REGISTRY.values():
+        pid = id(profile)
+        if pid not in seen:
+            seen.add(pid)
+            result.append(profile)
+    return result
+
+
+def _user_plugins_dir() -> Path | None:
+    """Return ``$HERMES_HOME/plugins/model-providers/`` if it exists."""
+    try:
+        from hermes_constants import get_hermes_home
+
+        d = get_hermes_home() / "plugins" / "model-providers"
+        return d if d.is_dir() else None
+    except Exception:
+        return None
+
+
+def _import_plugin_dir(plugin_dir: Path, source: str) -> None:
+    """Import a single plugin directory so it self-registers.
+
+    ``source`` is "bundled" or "user", used only for log messages.
+    """
+    init_file = plugin_dir / "__init__.py"
+    if not init_file.exists():
+        return
+
+    # Give bundled plugins a stable import path (``plugins.model_providers.<name>``)
+    # so relative imports within the plugin work. User plugins load via
+    # ``importlib.util.spec_from_file_location`` with a unique module name so
+    # multiple HERMES_HOME profiles don't alias each other.
+    safe_name = plugin_dir.name.replace("-", "_")
+    if source == "bundled":
+        module_name = f"plugins.model_providers.{safe_name}"
+    else:
+        module_name = f"_hermes_user_provider_{safe_name}"
+
+    if module_name in sys.modules:
+        return  # already imported
+
+    try:
+        spec = importlib.util.spec_from_file_location(
+            module_name, init_file, submodule_search_locations=[str(plugin_dir)]
+        )
+        if spec is None or spec.loader is None:
+            return
+        module = importlib.util.module_from_spec(spec)
+        sys.modules[module_name] = module
+        spec.loader.exec_module(module)
+    except Exception as exc:
+        logger.warning(
+            "Failed to load %s provider plugin %s: %s", source, plugin_dir.name, exc
+        )
+        sys.modules.pop(module_name, None)
+
+
+def _discover_providers() -> None:
+    """Populate the registry by importing every provider plugin.
+
+    Order:
+      1. Bundled plugins at ``<repo>/plugins/model-providers/<name>/``
+      2. User plugins at ``$HERMES_HOME/plugins/model-providers/<name>/``
+      3. Legacy per-file modules at ``providers/<name>.py`` (back-compat)
+
+    Each step imports its plugins, which call ``register_provider()`` at
+    module-level. Later steps win on name collision.
+    """
+    global _discovered
+    if _discovered:
+        return
+    _discovered = True
+
+    # 1. Bundled plugins — shipped with hermes-agent.
+    if _BUNDLED_PLUGINS_DIR.is_dir():
+        for child in sorted(_BUNDLED_PLUGINS_DIR.iterdir()):
+            if not child.is_dir() or child.name.startswith(("_", ".")):
+                continue
+            _import_plugin_dir(child, "bundled")
+
+    # 2. User plugins — under $HERMES_HOME/plugins/model-providers/<name>/.
+    #    These can override any bundled profile of the same name (last-writer-wins
+    #    in register_provider()).
+    user_dir = _user_plugins_dir()
+    if user_dir is not None:
+        for child in sorted(user_dir.iterdir()):
+            if not child.is_dir() or child.name.startswith(("_", ".")):
+                continue
+            _import_plugin_dir(child, "user")
+
+    # 3. Legacy single-file profiles at providers/<name>.py. Kept for
+    #    back-compat — if someone drops a ``providers/foo.py`` into an
+    #    editable install, it still works without the plugin layout.
+    try:
+        import pkgutil
+
+        import providers as _pkg
+
+        for _importer, modname, _ispkg in pkgutil.iter_modules(_pkg.__path__):
+            if modname.startswith("_") or modname == "base":
+                continue
+            try:
+                importlib.import_module(f"providers.{modname}")
+            except ImportError as exc:
+                logger.warning(
+                    "Failed to import legacy provider module %s: %s", modname, exc
+                )
+    except Exception:
+        pass

providers/base.py

@@ -0,0 +1,165 @@
+"""Provider profile base class.
+
+A ProviderProfile declares everything about an inference provider in one place:
+auth, endpoints, client quirks, request-time quirks. The transport reads this
+instead of receiving 20+ boolean flags.
+
+Provider profiles are DECLARATIVE — they describe the provider's behavior.
+They do NOT own client construction, credential rotation, or streaming.
+Those stay on AIAgent.
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass, field
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+# Sentinel for "omit temperature entirely" (Kimi: server manages it)
+OMIT_TEMPERATURE = object()
+
+
+@dataclass
+class ProviderProfile:
+    """Base provider profile — subclass or instantiate with overrides."""
+
+    # ── Identity ─────────────────────────────────────────────
+    name: str
+    api_mode: str = "chat_completions"
+    aliases: tuple = ()
+
+    # ── Human-readable metadata ───────────────────────────────
+    display_name: str = ""       # e.g. "GMI Cloud" — shown in picker/labels
+    description: str = ""        # e.g. "GMI Cloud (multi-model direct API)" — picker subtitle
+    signup_url: str = ""         # e.g. "https://www.gmicloud.ai/" — shown during setup
+
+    # ── Auth & endpoints ─────────────────────────────────────
+    env_vars: tuple = ()
+    base_url: str = ""
+    models_url: str = ""  # explicit models endpoint; falls back to {base_url}/models
+    auth_type: str = "api_key"   # api_key|oauth_device_code|oauth_external|copilot|aws_sdk
+
+    # ── Model catalog ─────────────────────────────────────────
+    # fallback_models: curated list shown in /model picker when live fetch fails.
+    # Only agentic models that support tool calling should appear here.
+    fallback_models: tuple = ()
+
+    # hostname: base hostname for URL→provider reverse-mapping in model_metadata.py
+    # e.g. "api.gmi-serving.com". Derived from base_url when empty.
+    hostname: str = ""
+
+    # ── Client-level quirks (set once at client construction) ─
+    default_headers: dict[str, str] = field(default_factory=dict)
+
+    # ── Request-level quirks ─────────────────────────────────
+    # Temperature: None = use caller's default, OMIT_TEMPERATURE = don't send
+    fixed_temperature: Any = None
+    default_max_tokens: int | None = None
+    default_aux_model: str = (
+        ""  # cheap model for auxiliary tasks (compression, vision, etc.)
+    )
+    # empty = use main model
+
+    # ── Hooks (override in subclass for complex providers) ───
+
+    def get_hostname(self) -> str:
+        """Return the provider's base hostname for URL-based detection.
+
+        Uses self.hostname if set explicitly, otherwise derives it from base_url.
+        e.g. 'https://api.gmi-serving.com/v1' → 'api.gmi-serving.com'
+        """
+        if self.hostname:
+            return self.hostname
+        if self.base_url:
+            from urllib.parse import urlparse
+            return urlparse(self.base_url).hostname or ""
+        return ""
+
+    def prepare_messages(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Provider-specific message preprocessing.
+
+        Called AFTER codex field sanitization, BEFORE developer role swap.
+        Default: pass-through.
+        """
+        return messages
+
+    def build_extra_body(
+        self, *, session_id: str | None = None, **context: Any
+    ) -> dict[str, Any]:
+        """Provider-specific extra_body fields.
+
+        Merged into the API kwargs extra_body. Default: empty dict.
+        """
+        return {}
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        **context: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        """Provider-specific kwargs split between extra_body and top-level api_kwargs.
+
+        Returns (extra_body_additions, top_level_kwargs).
+        The transport merges extra_body_additions into extra_body, and
+        top_level_kwargs directly into api_kwargs.
+
+        This split exists because some providers put reasoning config in
+        extra_body (OpenRouter: extra_body.reasoning) while others put it
+        as top-level api_kwargs (Kimi: api_kwargs.reasoning_effort).
+
+        Default: ({}, {}).
+        """
+        return {}, {}
+
+    def fetch_models(
+        self,
+        *,
+        api_key: str | None = None,
+        timeout: float = 8.0,
+    ) -> list[str] | None:
+        """Fetch the live model list from the provider's models endpoint.
+
+        Returns a list of model ID strings, or None if the fetch failed or
+        the provider does not support live model listing.
+
+        Resolution order for the endpoint URL:
+          1. self.models_url  (explicit override — use when the models
+             endpoint differs from the inference base URL, e.g. OpenRouter
+             exposes a public catalog at /api/v1/models while inference is
+             at /api/v1)
+          2. self.base_url + "/models"  (standard OpenAI-compat fallback)
+
+        The default implementation sends Bearer auth when api_key is given
+        and forwards self.default_headers. Override to customise auth, path,
+        response shape, or to return None for providers with no REST catalog.
+
+        Callers must always fall back to the static _PROVIDER_MODELS list
+        when this returns None.
+        """
+        url = (self.models_url or "").strip()
+        if not url:
+            if not self.base_url:
+                return None
+            url = self.base_url.rstrip("/") + "/models"
+
+        import json
+        import urllib.request
+
+        req = urllib.request.Request(url)
+        if api_key:
+            req.add_header("Authorization", f"Bearer {api_key}")
+        req.add_header("Accept", "application/json")
+        for k, v in self.default_headers.items():
+            req.add_header(k, v)
+
+        try:
+            with urllib.request.urlopen(req, timeout=timeout) as resp:
+                data = json.loads(resp.read().decode())
+            items = data if isinstance(data, list) else data.get("data", [])
+            return [m["id"] for m in items if isinstance(m, dict) and "id" in m]
+        except Exception as exc:
+            logger.debug("fetch_models(%s): %s", self.name, exc)
+            return None

pyproject.toml

@@ -142,7 +142,7 @@ hermes_cli = ["web_dist/**/*"]
 gateway = ["assets/**/*"]
 
 [tool.setuptools.packages.find]
-include = ["agent", "agent.*", "tools", "tools.*", "hermes_cli", "gateway", "gateway.*", "tui_gateway", "tui_gateway.*", "cron", "acp_adapter", "plugins", "plugins.*"]
+include = ["agent", "agent.*", "tools", "tools.*", "hermes_cli", "gateway", "gateway.*", "tui_gateway", "tui_gateway.*", "cron", "acp_adapter", "plugins", "plugins.*", "providers", "providers.*"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]

run_agent.py

@@ -1451,6 +1451,17 @@ class AIAgent:
                 elif base_url_host_matches(effective_base, "chatgpt.com"):
                     from agent.auxiliary_client import _codex_cloudflare_headers
                     client_kwargs["default_headers"] = _codex_cloudflare_headers(api_key)
+                elif "default_headers" not in client_kwargs:
+                    # Fall back to profile.default_headers for providers that
+                    # declare custom headers (e.g. Vercel AI Gateway attribution,
+                    # Kimi User-Agent on non-kimi.com endpoints).
+                    try:
+                        from providers import get_provider_profile as _gpf
+                        _ph = _gpf(self.provider)
+                        if _ph and _ph.default_headers:
+                            client_kwargs["default_headers"] = dict(_ph.default_headers)
+                    except Exception:
+                        pass
             else:
                 # No explicit creds — use the centralized provider router
                 from agent.auxiliary_client import resolve_provider_client
@@ -6251,7 +6262,19 @@ class AIAgent:
                 self._client_kwargs.get("api_key", "")
             )
         else:
-            self._client_kwargs.pop("default_headers", None)
+            # No URL-specific headers — check profile.default_headers before clearing.
+            _ph_headers = None
+            try:
+                from providers import get_provider_profile as _gpf2
+                _ph2 = _gpf2(self.provider)
+                if _ph2 and _ph2.default_headers:
+                    _ph_headers = dict(_ph2.default_headers)
+            except Exception:
+                pass
+            if _ph_headers:
+                self._client_kwargs["default_headers"] = _ph_headers
+            else:
+                self._client_kwargs.pop("default_headers", None)
 
     def _swap_credential(self, entry) -> None:
         runtime_key = getattr(entry, "runtime_api_key", None) or getattr(entry, "access_token", "")
@@ -8469,7 +8492,7 @@ class AIAgent:
             _omit_temp = False
             _fixed_temp = None
 
-        # Provider preferences (OpenRouter-specific)
+        # Provider preferences (OpenRouter-style)
         _prefs: Dict[str, Any] = {}
         if self.providers_allowed:
             _prefs["only"] = self.providers_allowed
@@ -8484,16 +8507,16 @@ class AIAgent:
         if self.provider_data_collection:
             _prefs["data_collection"] = self.provider_data_collection
 
-        # Anthropic max output for Claude on OpenRouter/Nous
+        # Claude max-output override on aggregators
         _ant_max = None
         if (_is_or or _is_nous) and "claude" in (self.model or "").lower():
             try:
                 from agent.anthropic_adapter import _get_anthropic_max_output
                 _ant_max = _get_anthropic_max_output(self.model)
             except Exception:
-                pass  # fail open — let the proxy pick its default
+                pass
 
-        # Qwen session metadata precomputed here (promptId is per-call random)
+        # Qwen session metadata
         _qwen_meta = None
         if _is_qwen:
             _qwen_meta = {
@@ -8501,8 +8524,44 @@ class AIAgent:
                 "promptId": str(uuid.uuid4()),
             }
 
-        # Ephemeral max output override — consume immediately so the next
-        # turn doesn't inherit it.
+        # ── Provider profile path (registered providers) ───────────────────
+        # Profiles handle per-provider quirks via hooks. When a profile is
+        # found, delegate fully; otherwise fall through to the legacy flag path.
+        try:
+            from providers import get_provider_profile
+            _profile = get_provider_profile(self.provider)
+        except Exception:
+            _profile = None
+
+        if _profile:
+            _ephemeral_out = getattr(self, "_ephemeral_max_output_tokens", None)
+            if _ephemeral_out is not None:
+                self._ephemeral_max_output_tokens = None
+
+            return _ct.build_kwargs(
+                model=self.model,
+                messages=api_messages,
+                tools=self.tools,
+                base_url=self.base_url,
+                timeout=self._resolved_api_call_timeout(),
+                max_tokens=self.max_tokens,
+                ephemeral_max_output_tokens=_ephemeral_out,
+                max_tokens_param_fn=self._max_tokens_param,
+                reasoning_config=self.reasoning_config,
+                request_overrides=self.request_overrides,
+                session_id=getattr(self, "session_id", None),
+                provider_profile=_profile,
+                ollama_num_ctx=self._ollama_num_ctx,
+                # Context forwarded to profile hooks:
+                provider_preferences=_prefs or None,
+                anthropic_max_output=_ant_max,
+                supports_reasoning=self._supports_reasoning_extra_body(),
+                qwen_session_metadata=_qwen_meta,
+            )
+
+        # ── Legacy flag path ────────────────────────────────────────────
+        # Reached only when get_provider_profile() returns None — i.e. a
+        # completely unknown provider not in providers/ registry.
         _ephemeral_out = getattr(self, "_ephemeral_max_output_tokens", None)
         if _ephemeral_out is not None:
             self._ephemeral_max_output_tokens = None

tests/agent/test_minimax_provider.py

@@ -71,17 +71,17 @@ class TestMinimaxThinkingSupport:
 
 
 class TestMinimaxAuxModel:
-    """Verify auxiliary model is standard (not highspeed)."""
+    """Verify auxiliary model is standard (not highspeed) — now reads from profiles."""
 
     def test_minimax_aux_is_standard(self):
-        from agent.auxiliary_client import _API_KEY_PROVIDER_AUX_MODELS
-        assert _API_KEY_PROVIDER_AUX_MODELS["minimax"] == "MiniMax-M2.7"
-        assert _API_KEY_PROVIDER_AUX_MODELS["minimax-cn"] == "MiniMax-M2.7"
+        from agent.auxiliary_client import _get_aux_model_for_provider
+        assert _get_aux_model_for_provider("minimax") == "MiniMax-M2.7"
+        assert _get_aux_model_for_provider("minimax-cn") == "MiniMax-M2.7"
 
     def test_minimax_aux_not_highspeed(self):
-        from agent.auxiliary_client import _API_KEY_PROVIDER_AUX_MODELS
-        assert "highspeed" not in _API_KEY_PROVIDER_AUX_MODELS["minimax"]
-        assert "highspeed" not in _API_KEY_PROVIDER_AUX_MODELS["minimax-cn"]
+        from agent.auxiliary_client import _get_aux_model_for_provider
+        assert "highspeed" not in _get_aux_model_for_provider("minimax")
+        assert "highspeed" not in _get_aux_model_for_provider("minimax-cn")
 
 
 class TestMinimaxBetaHeaders:

tests/agent/transports/test_chat_completions.py

@@ -73,17 +73,21 @@ class TestChatCompletionsBuildKwargs:
         assert kw["tools"] == tools
 
     def test_openrouter_provider_prefs(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("openrouter")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
             model="gpt-4o", messages=msgs,
-            is_openrouter=True,
+            provider_profile=profile,
             provider_preferences={"only": ["openai"]},
         )
         assert kw["extra_body"]["provider"] == {"only": ["openai"]}
 
     def test_nous_tags(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("nous")
         msgs = [{"role": "user", "content": "Hi"}]
-        kw = transport.build_kwargs(model="gpt-4o", messages=msgs, is_nous=True)
+        kw = transport.build_kwargs(model="gpt-4o", messages=msgs, provider_profile=profile)
         assert kw["extra_body"]["tags"] == ["product=hermes-agent"]
 
     def test_reasoning_default(self, transport):
@@ -95,29 +99,36 @@ class TestChatCompletionsBuildKwargs:
         assert kw["extra_body"]["reasoning"] == {"enabled": True, "effort": "medium"}
 
     def test_nous_omits_disabled_reasoning(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("nous")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
             model="gpt-4o", messages=msgs,
+            provider_profile=profile,
             supports_reasoning=True,
-            is_nous=True,
             reasoning_config={"enabled": False},
         )
         # Nous rejects enabled=false; reasoning omitted entirely
         assert "reasoning" not in kw.get("extra_body", {})
 
     def test_ollama_num_ctx(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("custom")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
             model="llama3", messages=msgs,
+            provider_profile=profile,
             ollama_num_ctx=32768,
         )
         assert kw["extra_body"]["options"]["num_ctx"] == 32768
 
     def test_custom_think_false(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("custom")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
             model="qwen3", messages=msgs,
-            is_custom_provider=True,
+            provider_profile=profile,
             reasoning_config={"effort": "none"},
         )
         assert kw["extra_body"]["think"] is False
@@ -304,23 +315,29 @@ class TestChatCompletionsBuildKwargs:
         assert kw["max_tokens"] == 2048
 
     def test_nvidia_default_max_tokens(self, transport):
+        """NVIDIA max_tokens=16384 is now set via ProviderProfile, not legacy flag."""
+        from providers import get_provider_profile
+
+        profile = get_provider_profile("nvidia")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
-            model="glm-4.7", messages=msgs,
-            is_nvidia_nim=True,
+            model="nvidia/llama-3.1-405b-instruct",
+            messages=msgs,
             max_tokens_param_fn=lambda n: {"max_tokens": n},
+            provider_profile=profile,
         )
-        # NVIDIA default: 16384
         assert kw["max_tokens"] == 16384
 
     def test_qwen_default_max_tokens(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("qwen-oauth")
         msgs = [{"role": "user", "content": "Hi"}]
         kw = transport.build_kwargs(
             model="qwen3-coder-plus", messages=msgs,
-            is_qwen_portal=True,
+            provider_profile=profile,
             max_tokens_param_fn=lambda n: {"max_tokens": n},
         )
-        # Qwen default: 65536
+        # Qwen default: 65536 from profile.default_max_tokens
         assert kw["max_tokens"] == 65536
 
     def test_anthropic_max_output_for_claude_on_aggregator(self, transport):
@@ -343,14 +360,23 @@ class TestChatCompletionsBuildKwargs:
         assert kw["service_tier"] == "priority"
 
     def test_fixed_temperature(self, transport):
+        """Fixed temperature is now set via ProviderProfile.fixed_temperature."""
+        from providers.base import ProviderProfile
         msgs = [{"role": "user", "content": "Hi"}]
-        kw = transport.build_kwargs(model="gpt-4o", messages=msgs, fixed_temperature=0.6)
+        kw = transport.build_kwargs(
+            model="gpt-4o", messages=msgs,
+            provider_profile=ProviderProfile(name="_t", fixed_temperature=0.6),
+        )
         assert kw["temperature"] == 0.6
 
     def test_omit_temperature(self, transport):
+        """Omit temperature is set via ProviderProfile with OMIT_TEMPERATURE sentinel."""
+        from providers.base import ProviderProfile, OMIT_TEMPERATURE
         msgs = [{"role": "user", "content": "Hi"}]
-        kw = transport.build_kwargs(model="gpt-4o", messages=msgs, omit_temperature=True, fixed_temperature=0.5)
-        # omit wins
+        kw = transport.build_kwargs(
+            model="gpt-4o", messages=msgs,
+            provider_profile=ProviderProfile(name="_t", fixed_temperature=OMIT_TEMPERATURE),
+        )
         assert "temperature" not in kw
 
 
@@ -358,18 +384,22 @@ class TestChatCompletionsKimi:
     """Regression tests for the Kimi/Moonshot quirks migrated into the transport."""
 
     def test_kimi_max_tokens_default(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("kimi-coding")
         kw = transport.build_kwargs(
             model="kimi-k2", messages=[{"role": "user", "content": "Hi"}],
-            is_kimi=True,
+            provider_profile=profile,
             max_tokens_param_fn=lambda n: {"max_tokens": n},
         )
-        # Kimi CLI default: 32000
+        # Kimi CLI default: 32000 from KimiProfile.default_max_tokens
         assert kw["max_tokens"] == 32000
 
     def test_kimi_reasoning_effort_top_level(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("kimi-coding")
         kw = transport.build_kwargs(
             model="kimi-k2", messages=[{"role": "user", "content": "Hi"}],
-            is_kimi=True,
+            provider_profile=profile,
             reasoning_config={"effort": "high"},
             max_tokens_param_fn=lambda n: {"max_tokens": n},
         )
@@ -387,17 +417,21 @@ class TestChatCompletionsKimi:
         assert "reasoning_effort" not in kw
 
     def test_kimi_thinking_enabled_extra_body(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("kimi-coding")
         kw = transport.build_kwargs(
             model="kimi-k2", messages=[{"role": "user", "content": "Hi"}],
-            is_kimi=True,
+            provider_profile=profile,
             max_tokens_param_fn=lambda n: {"max_tokens": n},
         )
         assert kw["extra_body"]["thinking"] == {"type": "enabled"}
 
     def test_kimi_thinking_disabled_extra_body(self, transport):
+        from providers import get_provider_profile
+        profile = get_provider_profile("kimi-coding")
         kw = transport.build_kwargs(
             model="kimi-k2", messages=[{"role": "user", "content": "Hi"}],
-            is_kimi=True,
+            provider_profile=profile,
             reasoning_config={"enabled": False},
             max_tokens_param_fn=lambda n: {"max_tokens": n},
         )

tests/hermes_cli/test_gmi_provider.py

@@ -269,9 +269,9 @@ class TestGmiModelMetadata:
 
 class TestGmiAuxiliary:
     def test_aux_default_model(self):
-        from agent.auxiliary_client import _API_KEY_PROVIDER_AUX_MODELS
+        from agent.auxiliary_client import _get_aux_model_for_provider
 
-        assert _API_KEY_PROVIDER_AUX_MODELS["gmi"] == "google/gemini-3.1-flash-lite-preview"
+        assert _get_aux_model_for_provider("gmi") == "google/gemini-3.1-flash-lite-preview"
 
     def test_resolve_provider_client_uses_gmi_aux_default(self, monkeypatch):
         monkeypatch.setenv("GMI_API_KEY", "gmi-test-key")

tests/providers/test_e2e_wiring.py

@@ -0,0 +1,118 @@
+"""E2E tests: verify _build_kwargs_from_profile produces correct output.
+
+These tests call _build_kwargs_from_profile on the transport directly,
+without importing run_agent (which would cause xdist worker contamination).
+"""
+
+import pytest
+from agent.transports.chat_completions import ChatCompletionsTransport
+from providers import get_provider_profile
+
+
+@pytest.fixture
+def transport():
+    return ChatCompletionsTransport()
+
+
+def _msgs():
+    return [{"role": "user", "content": "hi"}]
+
+
+class TestNvidiaProfileWiring:
+    def test_nvidia_gets_default_max_tokens(self, transport):
+        profile = get_provider_profile("nvidia")
+        kwargs = transport.build_kwargs(
+            model="nvidia/llama-3.1-nemotron-70b-instruct",
+            messages=_msgs(),
+            tools=None,
+            provider_profile=profile,
+            max_tokens=None,
+            max_tokens_param_fn=lambda x: {"max_tokens": x} if x else {},
+            timeout=300,
+            reasoning_config=None,
+            request_overrides=None,
+            session_id="test",
+            ollama_num_ctx=None,
+        )
+        # NVIDIA profile sets default_max_tokens=16384
+        assert kwargs.get("max_tokens") == 16384
+
+    def test_nvidia_nim_alias(self, transport):
+        profile = get_provider_profile("nvidia-nim")
+        assert profile is not None
+        assert profile.name == "nvidia"
+        assert profile.default_max_tokens == 16384
+
+    def test_nvidia_model_passed(self, transport):
+        profile = get_provider_profile("nvidia")
+        kwargs = transport.build_kwargs(
+            model="nvidia/test-model",
+            messages=_msgs(),
+            tools=None,
+            provider_profile=profile,
+            max_tokens=None,
+            max_tokens_param_fn=lambda x: {"max_tokens": x} if x else {},
+            timeout=300,
+            reasoning_config=None,
+            request_overrides=None,
+            session_id="test",
+            ollama_num_ctx=None,
+        )
+        assert kwargs["model"] == "nvidia/test-model"
+
+    def test_nvidia_messages_passed(self, transport):
+        profile = get_provider_profile("nvidia")
+        msgs = _msgs()
+        kwargs = transport.build_kwargs(
+            model="nvidia/test",
+            messages=msgs,
+            tools=None,
+            provider_profile=profile,
+            max_tokens=None,
+            max_tokens_param_fn=lambda x: {"max_tokens": x} if x else {},
+            timeout=300,
+            reasoning_config=None,
+            request_overrides=None,
+            session_id="test",
+            ollama_num_ctx=None,
+        )
+        assert kwargs["messages"] == msgs
+
+
+class TestDeepSeekProfileWiring:
+    def test_deepseek_no_forced_max_tokens(self, transport):
+        profile = get_provider_profile("deepseek")
+        kwargs = transport.build_kwargs(
+            model="deepseek-chat",
+            messages=_msgs(),
+            tools=None,
+            provider_profile=profile,
+            max_tokens=None,
+            max_tokens_param_fn=lambda x: {"max_tokens": x} if x else {},
+            timeout=300,
+            reasoning_config=None,
+            request_overrides=None,
+            session_id="test",
+            ollama_num_ctx=None,
+        )
+        # DeepSeek has no default_max_tokens
+        assert kwargs["model"] == "deepseek-chat"
+        assert kwargs.get("max_tokens") is None or "max_tokens" not in kwargs
+
+    def test_deepseek_messages_passed(self, transport):
+        profile = get_provider_profile("deepseek")
+        msgs = _msgs()
+        kwargs = transport.build_kwargs(
+            model="deepseek-chat",
+            messages=msgs,
+            tools=None,
+            provider_profile=profile,
+            max_tokens=None,
+            max_tokens_param_fn=lambda x: {"max_tokens": x} if x else {},
+            timeout=300,
+            reasoning_config=None,
+            request_overrides=None,
+            session_id="test",
+            ollama_num_ctx=None,
+        )
+        assert kwargs["messages"] == msgs

tests/providers/test_plugin_discovery.py

@@ -0,0 +1,145 @@
+"""Tests for the model-providers plugin discovery system.
+
+Verifies that:
+ 1. All bundled providers at plugins/model-providers/<name>/ are discovered
+ 2. User plugins at $HERMES_HOME/plugins/model-providers/<name>/ override bundled
+ 3. plugin.yaml manifests with kind=model-provider are correctly categorized
+"""
+
+from __future__ import annotations
+
+import importlib
+import sys
+from pathlib import Path
+
+import pytest
+
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+
+
+def _clear_provider_caches():
+    """Force providers/__init__.py to re-discover on next list_providers()."""
+    import providers as _pkg
+    _pkg._REGISTRY.clear()
+    _pkg._ALIASES.clear()
+    _pkg._discovered = False
+    # Evict any cached plugin modules so the next import re-executes.
+    for mod in list(sys.modules.keys()):
+        if (
+            mod.startswith("plugins.model_providers")
+            or mod.startswith("_hermes_user_provider")
+        ):
+            del sys.modules[mod]
+
+
+def test_bundled_plugins_discovered():
+    """Every plugins/model-providers/<name>/ should contain a plugin.yaml + __init__.py."""
+    plugins_dir = REPO_ROOT / "plugins" / "model-providers"
+    assert plugins_dir.is_dir(), f"Missing {plugins_dir}"
+
+    child_dirs = [c for c in plugins_dir.iterdir() if c.is_dir()]
+    assert len(child_dirs) >= 28, f"Expected at least 28 provider plugins, found {len(child_dirs)}"
+
+    for child in child_dirs:
+        assert (child / "__init__.py").exists(), f"{child.name} missing __init__.py"
+        assert (child / "plugin.yaml").exists(), f"{child.name} missing plugin.yaml"
+
+
+def test_all_33_profiles_register():
+    """After discovery, the registry must contain exactly 33 distinct profiles."""
+    _clear_provider_caches()
+    from providers import list_providers
+
+    profiles = list_providers()
+    names = sorted(p.name for p in profiles)
+    assert len(names) == 33, f"Expected 33 profiles, got {len(names)}: {names}"
+
+    # Spot-check representative providers from different categories
+    for required in (
+        "openrouter", "anthropic", "custom", "bedrock", "openai-codex",
+        "minimax-oauth", "gmi", "xiaomi", "alibaba-coding-plan",
+    ):
+        assert required in names, f"Missing profile: {required}"
+
+
+def test_user_plugin_overrides_bundled(tmp_path, monkeypatch):
+    """A user plugin with the same name must override the bundled profile."""
+    # Point HERMES_HOME at a fresh temp dir
+    hermes_home = tmp_path / ".hermes"
+    hermes_home.mkdir()
+    monkeypatch.setenv("HERMES_HOME", str(hermes_home))
+    # get_hermes_home() may be module-cached depending on codebase; ensure the
+    # env var is the source of truth. Most code paths re-read it each call.
+
+    # Drop a user plugin that replaces 'gmi'
+    user_gmi = hermes_home / "plugins" / "model-providers" / "gmi"
+    user_gmi.mkdir(parents=True)
+    (user_gmi / "__init__.py").write_text(
+        "from providers import register_provider\n"
+        "from providers.base import ProviderProfile\n"
+        "\n"
+        "custom_gmi = ProviderProfile(\n"
+        '    name="gmi",\n'
+        '    aliases=("gmi-user-override-test",),\n'
+        '    env_vars=("GMI_API_KEY",),\n'
+        '    base_url="https://user-override.example.com/v1",\n'
+        '    auth_type="api_key",\n'
+        ")\n"
+        "register_provider(custom_gmi)\n"
+    )
+    (user_gmi / "plugin.yaml").write_text(
+        "name: gmi-user-override\n"
+        "kind: model-provider\n"
+        "version: 0.0.1\n"
+        "description: Test user override\n"
+    )
+
+    _clear_provider_caches()
+    from providers import get_provider_profile
+
+    gmi = get_provider_profile("gmi")
+    assert gmi is not None
+    assert gmi.base_url == "https://user-override.example.com/v1", (
+        f"User override not applied; got base_url={gmi.base_url!r}"
+    )
+    assert "gmi-user-override-test" in gmi.aliases
+
+    # Clean up: reset discovery state so other tests see the bundled version
+    _clear_provider_caches()
+
+
+def test_general_plugin_manager_skips_model_provider_kind(tmp_path, monkeypatch):
+    """The general PluginManager must NOT import model-provider plugins
+    (providers/__init__.py handles them). It records the manifest only."""
+    from hermes_cli import plugins as plugin_mod
+
+    hermes_home = tmp_path / ".hermes"
+    hermes_home.mkdir()
+    monkeypatch.setenv("HERMES_HOME", str(hermes_home))
+
+    # Create a user-installed plugin with an explicit kind: model-provider.
+    user_plugin = hermes_home / "plugins" / "test-model-provider"
+    user_plugin.mkdir(parents=True)
+    (user_plugin / "plugin.yaml").write_text(
+        "name: test-model-provider\n"
+        "kind: model-provider\n"
+        "version: 0.0.1\n"
+    )
+    (user_plugin / "__init__.py").write_text(
+        # Intentionally broken import — if the general loader tries to
+        # import this module, the test will fail with ImportError.
+        "raise AssertionError('model-provider plugins must not be imported by PluginManager')\n"
+    )
+
+    # Fresh manager
+    manager = plugin_mod.PluginManager()
+    manager.discover_and_load(force=True)
+
+    # The manifest should be recorded but not loaded
+    loaded = manager._plugins.get("test-model-provider")
+    assert loaded is not None
+    assert loaded.manifest.kind == "model-provider"
+    # No import means the module must NOT be in the plugins list as a loaded one.
+    # We check that the general loader didn't crash and didn't raise from the
+    # broken __init__.py.

tests/providers/test_profile_wiring.py

@@ -0,0 +1,290 @@
+"""Profile-path parity tests: verify profile path produces identical output to legacy flags.
+
+Each test calls build_kwargs twice — once with legacy flags, once with provider_profile —
+and asserts the output is identical. This catches any behavioral drift between the two paths.
+"""
+
+import pytest
+from agent.transports.chat_completions import ChatCompletionsTransport
+from providers import get_provider_profile
+
+
+@pytest.fixture
+def transport():
+    return ChatCompletionsTransport()
+
+
+def _msgs():
+    return [{"role": "user", "content": "hello"}]
+
+
+def _max_tokens_fn(n):
+    return {"max_completion_tokens": n}
+
+
+class TestNvidiaProfileParity:
+    def test_max_tokens_match(self, transport):
+        """NVIDIA profile sets max_tokens=16384; legacy flag is removed."""
+        profile = transport.build_kwargs(
+            model="nvidia/nemotron", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("nvidia"),
+            max_tokens_param_fn=_max_tokens_fn,
+        )
+        assert profile["max_completion_tokens"] == 16384
+
+
+class TestKimiProfileParity:
+    def test_temperature_omitted(self, transport):
+        legacy = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi-coding"), omit_temperature=True,
+        )
+        profile = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi"),
+        )
+        assert "temperature" not in legacy
+        assert "temperature" not in profile
+
+    def test_max_tokens(self, transport):
+        legacy = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi-coding"), max_tokens_param_fn=_max_tokens_fn,
+        )
+        profile = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi"),
+            max_tokens_param_fn=_max_tokens_fn,
+        )
+        assert profile["max_completion_tokens"] == legacy["max_completion_tokens"] == 32000
+
+    def test_thinking_enabled(self, transport):
+        rc = {"enabled": True, "effort": "high"}
+        legacy = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi-coding"), reasoning_config=rc,
+        )
+        profile = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi"),
+            reasoning_config=rc,
+        )
+        assert profile["extra_body"]["thinking"] == legacy["extra_body"]["thinking"]
+        assert profile["reasoning_effort"] == legacy["reasoning_effort"] == "high"
+
+    def test_thinking_disabled(self, transport):
+        rc = {"enabled": False}
+        legacy = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi-coding"), reasoning_config=rc,
+        )
+        profile = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi"),
+            reasoning_config=rc,
+        )
+        assert profile["extra_body"]["thinking"] == legacy["extra_body"]["thinking"]
+        assert profile["extra_body"]["thinking"]["type"] == "disabled"
+        assert "reasoning_effort" not in profile
+        assert "reasoning_effort" not in legacy
+
+    def test_reasoning_effort_default(self, transport):
+        rc = {"enabled": True}
+        legacy = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi-coding"), reasoning_config=rc,
+        )
+        profile = transport.build_kwargs(
+            model="kimi-k2", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("kimi"),
+            reasoning_config=rc,
+        )
+        assert profile["reasoning_effort"] == legacy["reasoning_effort"] == "medium"
+
+
+class TestOpenRouterProfileParity:
+    def test_provider_preferences(self, transport):
+        prefs = {"allow": ["anthropic"]}
+        legacy = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"), provider_preferences=prefs,
+        )
+        profile = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            provider_preferences=prefs,
+        )
+        assert profile["extra_body"]["provider"] == legacy["extra_body"]["provider"]
+
+    def test_reasoning_full_config(self, transport):
+        rc = {"enabled": True, "effort": "high"}
+        legacy = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"), supports_reasoning=True, reasoning_config=rc,
+        )
+        profile = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            supports_reasoning=True, reasoning_config=rc,
+        )
+        assert profile["extra_body"]["reasoning"] == legacy["extra_body"]["reasoning"]
+
+    def test_default_reasoning(self, transport):
+        legacy = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"), supports_reasoning=True,
+        )
+        profile = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            supports_reasoning=True,
+        )
+        assert profile["extra_body"]["reasoning"] == legacy["extra_body"]["reasoning"]
+
+
+class TestNousProfileParity:
+    def test_tags(self, transport):
+        legacy = transport.build_kwargs(
+            model="hermes-3", messages=_msgs(), tools=None, provider_profile=get_provider_profile("nous"),
+        )
+        profile = transport.build_kwargs(
+            model="hermes-3", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("nous"),
+        )
+        assert profile["extra_body"]["tags"] == legacy["extra_body"]["tags"]
+
+    def test_reasoning_omitted_when_disabled(self, transport):
+        rc = {"enabled": False}
+        legacy = transport.build_kwargs(
+            model="hermes-3", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("nous"), supports_reasoning=True, reasoning_config=rc,
+        )
+        profile = transport.build_kwargs(
+            model="hermes-3", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("nous"),
+            supports_reasoning=True, reasoning_config=rc,
+        )
+        assert "reasoning" not in legacy.get("extra_body", {})
+        assert "reasoning" not in profile.get("extra_body", {})
+
+
+class TestQwenProfileParity:
+    def test_max_tokens(self, transport):
+        legacy = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("qwen-oauth"), max_tokens_param_fn=_max_tokens_fn,
+        )
+        profile = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("qwen"),
+            max_tokens_param_fn=_max_tokens_fn,
+        )
+        assert profile["max_completion_tokens"] == legacy["max_completion_tokens"] == 65536
+
+    def test_vl_high_resolution(self, transport):
+        legacy = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None, provider_profile=get_provider_profile("qwen-oauth"),
+        )
+        profile = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("qwen"),
+        )
+        assert profile["extra_body"]["vl_high_resolution_images"] == legacy["extra_body"]["vl_high_resolution_images"]
+
+    def test_metadata_top_level(self, transport):
+        meta = {"sessionId": "s123", "promptId": "p456"}
+        legacy = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("qwen-oauth"), qwen_session_metadata=meta,
+        )
+        profile = transport.build_kwargs(
+            model="qwen3.5", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("qwen"),
+            qwen_session_metadata=meta,
+        )
+        assert profile["metadata"] == legacy["metadata"] == meta
+        assert "metadata" not in profile.get("extra_body", {})
+
+    def test_message_preprocessing(self, transport):
+        """Qwen profile normalizes string content to list-of-parts."""
+        msgs = [
+            {"role": "system", "content": "You are helpful."},
+            {"role": "user", "content": "hello"},
+        ]
+        profile = transport.build_kwargs(
+            model="qwen3.5", messages=msgs, tools=None,
+            provider_profile=get_provider_profile("qwen"),
+        )
+        out_msgs = profile["messages"]
+        # System message content normalized + cache_control injected
+        assert isinstance(out_msgs[0]["content"], list)
+        assert out_msgs[0]["content"][0]["type"] == "text"
+        assert "cache_control" in out_msgs[0]["content"][-1]
+        # User message content normalized
+        assert isinstance(out_msgs[1]["content"], list)
+        assert out_msgs[1]["content"][0] == {"type": "text", "text": "hello"}
+
+
+class TestDeveloperRoleParity:
+    """Developer role swap must work on BOTH legacy and profile paths."""
+
+    def test_legacy_path_swaps_for_gpt5(self, transport):
+        msgs = [{"role": "system", "content": "Be helpful"}, {"role": "user", "content": "hi"}]
+        kw = transport.build_kwargs(
+            model="gpt-5.4", messages=msgs, tools=None,
+        )
+        assert kw["messages"][0]["role"] == "developer"
+
+    def test_profile_path_swaps_for_gpt5(self, transport):
+        msgs = [{"role": "system", "content": "Be helpful"}, {"role": "user", "content": "hi"}]
+        kw = transport.build_kwargs(
+            model="gpt-5.4", messages=msgs, tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+        )
+        assert kw["messages"][0]["role"] == "developer"
+
+    def test_profile_path_no_swap_for_claude(self, transport):
+        msgs = [{"role": "system", "content": "Be helpful"}, {"role": "user", "content": "hi"}]
+        kw = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6", messages=msgs, tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+        )
+        assert kw["messages"][0]["role"] == "system"
+
+
+class TestRequestOverridesParity:
+    """request_overrides with extra_body must merge identically on both paths."""
+
+    def test_extra_body_override_legacy(self, transport):
+        kw = transport.build_kwargs(
+            model="gpt-5.4", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            request_overrides={"extra_body": {"custom_key": "custom_val"}},
+        )
+        assert kw["extra_body"]["custom_key"] == "custom_val"
+
+    def test_extra_body_override_profile(self, transport):
+        kw = transport.build_kwargs(
+            model="gpt-5.4", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            request_overrides={"extra_body": {"custom_key": "custom_val"}},
+        )
+        assert kw["extra_body"]["custom_key"] == "custom_val"
+
+    def test_extra_body_override_merges_with_provider_body(self, transport):
+        """Override extra_body merges WITH provider extra_body, not replaces."""
+        kw = transport.build_kwargs(
+            model="hermes-3", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("nous"),
+            request_overrides={"extra_body": {"custom": True}},
+        )
+        assert kw["extra_body"]["tags"] == ["product=hermes-agent"]  # from profile
+        assert kw["extra_body"]["custom"] is True  # from override
+
+    def test_top_level_override(self, transport):
+        kw = transport.build_kwargs(
+            model="gpt-5.4", messages=_msgs(), tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            request_overrides={"top_p": 0.9},
+        )
+        assert kw["top_p"] == 0.9

tests/providers/test_provider_profiles.py

@@ -0,0 +1,203 @@
+"""Tests for the provider module registry and profiles."""
+
+import pytest
+from providers import get_provider_profile, _REGISTRY
+from providers.base import ProviderProfile, OMIT_TEMPERATURE
+
+
+class TestRegistry:
+    def test_discovery_populates_registry(self):
+        p = get_provider_profile("nvidia")
+        assert p is not None
+        assert p.name == "nvidia"
+
+    def test_alias_lookup(self):
+        assert get_provider_profile("kimi").name == "kimi-coding"
+        assert get_provider_profile("moonshot").name == "kimi-coding"
+        assert get_provider_profile("kimi-coding-cn").name == "kimi-coding-cn"
+        assert get_provider_profile("or").name == "openrouter"
+        assert get_provider_profile("nous-portal").name == "nous"
+        assert get_provider_profile("qwen").name == "qwen-oauth"
+        assert get_provider_profile("qwen-portal").name == "qwen-oauth"
+
+    def test_unknown_provider_returns_none(self):
+        assert get_provider_profile("nonexistent-provider") is None
+
+    def test_all_providers_have_name(self):
+        get_provider_profile("nvidia")  # trigger discovery
+        for name, profile in _REGISTRY.items():
+            assert profile.name == name
+
+
+class TestNvidiaProfile:
+    def test_max_tokens(self):
+        p = get_provider_profile("nvidia")
+        assert p.default_max_tokens == 16384
+
+    def test_no_special_temperature(self):
+        p = get_provider_profile("nvidia")
+        assert p.fixed_temperature is None
+
+    def test_base_url(self):
+        p = get_provider_profile("nvidia")
+        assert "nvidia.com" in p.base_url
+
+
+class TestKimiProfile:
+    def test_temperature_omit(self):
+        p = get_provider_profile("kimi")
+        assert p.fixed_temperature is OMIT_TEMPERATURE
+
+    def test_max_tokens(self):
+        p = get_provider_profile("kimi")
+        assert p.default_max_tokens == 32000
+
+    def test_cn_separate_profile(self):
+        p = get_provider_profile("kimi-coding-cn")
+        assert p.name == "kimi-coding-cn"
+        assert p.env_vars == ("KIMI_CN_API_KEY",)
+        assert "moonshot.cn" in p.base_url
+
+    def test_cn_not_alias_of_kimi(self):
+        kimi = get_provider_profile("kimi-coding")
+        cn = get_provider_profile("kimi-coding-cn")
+        assert kimi is not cn
+        assert kimi.base_url != cn.base_url
+
+    def test_thinking_enabled(self):
+        p = get_provider_profile("kimi")
+        eb, tl = p.build_api_kwargs_extras(reasoning_config={"enabled": True, "effort": "high"})
+        assert eb["thinking"] == {"type": "enabled"}
+        assert tl["reasoning_effort"] == "high"
+
+    def test_thinking_disabled(self):
+        p = get_provider_profile("kimi")
+        eb, tl = p.build_api_kwargs_extras(reasoning_config={"enabled": False})
+        assert eb["thinking"] == {"type": "disabled"}
+        assert "reasoning_effort" not in tl
+
+    def test_reasoning_effort_default(self):
+        p = get_provider_profile("kimi")
+        eb, tl = p.build_api_kwargs_extras(reasoning_config={"enabled": True})
+        assert tl["reasoning_effort"] == "medium"
+
+    def test_no_config_defaults(self):
+        p = get_provider_profile("kimi")
+        eb, tl = p.build_api_kwargs_extras(reasoning_config=None)
+        assert eb["thinking"] == {"type": "enabled"}
+        assert tl["reasoning_effort"] == "medium"
+
+
+class TestOpenRouterProfile:
+    def test_extra_body_with_prefs(self):
+        p = get_provider_profile("openrouter")
+        body = p.build_extra_body(provider_preferences={"allow": ["anthropic"]})
+        assert body["provider"] == {"allow": ["anthropic"]}
+
+    def test_extra_body_no_prefs(self):
+        p = get_provider_profile("openrouter")
+        body = p.build_extra_body()
+        assert body == {}
+
+    def test_reasoning_full_config(self):
+        p = get_provider_profile("openrouter")
+        eb, _ = p.build_api_kwargs_extras(
+            reasoning_config={"enabled": True, "effort": "high"},
+            supports_reasoning=True,
+        )
+        assert eb["reasoning"] == {"enabled": True, "effort": "high"}
+
+    def test_reasoning_disabled_still_passes(self):
+        """OpenRouter passes disabled reasoning through (unlike Nous)."""
+        p = get_provider_profile("openrouter")
+        eb, _ = p.build_api_kwargs_extras(
+            reasoning_config={"enabled": False},
+            supports_reasoning=True,
+        )
+        assert eb["reasoning"] == {"enabled": False}
+
+    def test_default_reasoning(self):
+        p = get_provider_profile("openrouter")
+        eb, _ = p.build_api_kwargs_extras(supports_reasoning=True)
+        assert eb["reasoning"] == {"enabled": True, "effort": "medium"}
+
+
+class TestNousProfile:
+    def test_tags(self):
+        p = get_provider_profile("nous")
+        body = p.build_extra_body()
+        assert body["tags"] == ["product=hermes-agent"]
+
+    def test_auth_type(self):
+        p = get_provider_profile("nous")
+        assert p.auth_type == "oauth_device_code"
+
+    def test_reasoning_enabled(self):
+        p = get_provider_profile("nous")
+        eb, _ = p.build_api_kwargs_extras(
+            reasoning_config={"enabled": True, "effort": "medium"},
+            supports_reasoning=True,
+        )
+        assert eb["reasoning"] == {"enabled": True, "effort": "medium"}
+
+    def test_reasoning_omitted_when_disabled(self):
+        p = get_provider_profile("nous")
+        eb, _ = p.build_api_kwargs_extras(
+            reasoning_config={"enabled": False},
+            supports_reasoning=True,
+        )
+        assert "reasoning" not in eb
+
+
+class TestQwenProfile:
+    def test_max_tokens(self):
+        p = get_provider_profile("qwen-oauth")
+        assert p.default_max_tokens == 65536
+
+    def test_auth_type(self):
+        p = get_provider_profile("qwen-oauth")
+        assert p.auth_type == "oauth_external"
+
+    def test_extra_body_vl(self):
+        p = get_provider_profile("qwen-oauth")
+        body = p.build_extra_body()
+        assert body["vl_high_resolution_images"] is True
+
+    def test_prepare_messages_normalizes_content(self):
+        p = get_provider_profile("qwen-oauth")
+        msgs = [
+            {"role": "system", "content": "Be helpful"},
+            {"role": "user", "content": "hello"},
+        ]
+        result = p.prepare_messages(msgs)
+        # System message: content normalized to list, cache_control on last part
+        assert isinstance(result[0]["content"], list)
+        assert result[0]["content"][-1].get("cache_control") == {"type": "ephemeral"}
+        assert result[0]["content"][-1]["text"] == "Be helpful"
+        # User message: content normalized to list
+        assert isinstance(result[1]["content"], list)
+        assert result[1]["content"][0]["text"] == "hello"
+
+    def test_metadata_top_level(self):
+        p = get_provider_profile("qwen-oauth")
+        meta = {"sessionId": "s123", "promptId": "p456"}
+        eb, tl = p.build_api_kwargs_extras(qwen_session_metadata=meta)
+        assert tl["metadata"] == meta
+        assert "metadata" not in eb
+
+
+class TestBaseProfile:
+    def test_prepare_messages_passthrough(self):
+        p = ProviderProfile(name="test")
+        msgs = [{"role": "user", "content": "hi"}]
+        assert p.prepare_messages(msgs) is msgs
+
+    def test_build_extra_body_empty(self):
+        p = ProviderProfile(name="test")
+        assert p.build_extra_body() == {}
+
+    def test_build_api_kwargs_extras_empty(self):
+        p = ProviderProfile(name="test")
+        eb, tl = p.build_api_kwargs_extras()
+        assert eb == {}
+        assert tl == {}

tests/providers/test_transport_parity.py

@@ -0,0 +1,258 @@
+"""Parity tests: pin the exact current transport behavior per provider.
+
+These tests document the flag-based contract between run_agent.py and
+ChatCompletionsTransport.build_kwargs(). When the next PR wires profiles
+to replace flags, every assertion here must still pass — any failure is
+a behavioral regression.
+"""
+
+import pytest
+from agent.transports.chat_completions import ChatCompletionsTransport
+from providers import get_provider_profile
+
+
+@pytest.fixture
+def transport():
+    return ChatCompletionsTransport()
+
+
+def _simple_messages():
+    return [{"role": "user", "content": "hello"}]
+
+
+def _max_tokens_fn(n):
+    return {"max_completion_tokens": n}
+
+
+class TestNvidiaParity:
+    """NVIDIA NIM: default max_tokens=16384."""
+
+    def test_default_max_tokens(self, transport):
+        """NVIDIA default max_tokens=16384 comes from profile, not legacy is_nvidia_nim flag."""
+        from providers import get_provider_profile
+
+        profile = get_provider_profile("nvidia")
+        kw = transport.build_kwargs(
+            model="nvidia/llama-3.1-nemotron-70b-instruct",
+            messages=_simple_messages(),
+            tools=None,
+            max_tokens_param_fn=_max_tokens_fn,
+            provider_profile=profile,
+        )
+        assert kw["max_completion_tokens"] == 16384
+
+    def test_user_max_tokens_overrides(self, transport):
+        from providers import get_provider_profile
+
+        profile = get_provider_profile("nvidia")
+        kw = transport.build_kwargs(
+            model="nvidia/llama-3.1-nemotron-70b-instruct",
+            messages=_simple_messages(),
+            tools=None,
+            max_tokens=4096,
+            max_tokens_param_fn=_max_tokens_fn,
+            provider_profile=profile,
+        )
+        assert kw["max_completion_tokens"] == 4096  # user overrides default
+
+
+class TestKimiParity:
+    """Kimi: OMIT temperature, max_tokens=32000, thinking + reasoning_effort."""
+
+    def test_temperature_omitted(self, transport):
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            omit_temperature=True,
+        )
+        assert "temperature" not in kw
+
+    def test_default_max_tokens(self, transport):
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            max_tokens_param_fn=_max_tokens_fn,
+        )
+        assert kw["max_completion_tokens"] == 32000
+
+    def test_thinking_enabled(self, transport):
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            reasoning_config={"enabled": True, "effort": "high"},
+        )
+        assert kw["extra_body"]["thinking"] == {"type": "enabled"}
+
+    def test_thinking_disabled(self, transport):
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            reasoning_config={"enabled": False},
+        )
+        assert kw["extra_body"]["thinking"] == {"type": "disabled"}
+
+    def test_reasoning_effort_top_level(self, transport):
+        """Kimi reasoning_effort is a TOP-LEVEL api_kwargs key, NOT in extra_body."""
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            reasoning_config={"enabled": True, "effort": "high"},
+        )
+        assert kw.get("reasoning_effort") == "high"
+        assert "reasoning_effort" not in kw.get("extra_body", {})
+
+    def test_reasoning_effort_default_medium(self, transport):
+        kw = transport.build_kwargs(
+            model="kimi-k2",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("kimi-coding"),
+            reasoning_config={"enabled": True},
+        )
+        assert kw.get("reasoning_effort") == "medium"
+
+
+class TestOpenRouterParity:
+    """OpenRouter: provider preferences, reasoning in extra_body."""
+
+    def test_provider_preferences(self, transport):
+        prefs = {"allow": ["anthropic"], "sort": "price"}
+        kw = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            provider_preferences=prefs,
+        )
+        assert kw["extra_body"]["provider"] == prefs
+
+    def test_reasoning_passes_full_config(self, transport):
+        """OpenRouter passes the FULL reasoning_config dict, not just effort."""
+        rc = {"enabled": True, "effort": "high"}
+        kw = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            supports_reasoning=True,
+            reasoning_config=rc,
+        )
+        assert kw["extra_body"]["reasoning"] == rc
+
+    def test_default_reasoning_when_no_config(self, transport):
+        """When supports_reasoning=True but no config, adds default."""
+        kw = transport.build_kwargs(
+            model="anthropic/claude-sonnet-4.6",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("openrouter"),
+            supports_reasoning=True,
+        )
+        assert kw["extra_body"]["reasoning"] == {"enabled": True, "effort": "medium"}
+
+
+class TestNousParity:
+    """Nous: product tags, reasoning, omit when disabled."""
+
+    def test_tags(self, transport):
+        kw = transport.build_kwargs(
+            model="hermes-3-llama-3.1-405b",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("nous"),
+        )
+        assert kw["extra_body"]["tags"] == ["product=hermes-agent"]
+
+    def test_reasoning_omitted_when_disabled(self, transport):
+        """Nous special case: reasoning omitted entirely when disabled."""
+        kw = transport.build_kwargs(
+            model="hermes-3-llama-3.1-405b",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("nous"),
+            supports_reasoning=True,
+            reasoning_config={"enabled": False},
+        )
+        assert "reasoning" not in kw.get("extra_body", {})
+
+    def test_reasoning_enabled(self, transport):
+        rc = {"enabled": True, "effort": "high"}
+        kw = transport.build_kwargs(
+            model="hermes-3-llama-3.1-405b",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("nous"),
+            supports_reasoning=True,
+            reasoning_config=rc,
+        )
+        assert kw["extra_body"]["reasoning"] == rc
+
+
+class TestQwenParity:
+    """Qwen: max_tokens=65536, vl_high_resolution, metadata top-level."""
+
+    def test_default_max_tokens(self, transport):
+        kw = transport.build_kwargs(
+            model="qwen3.5-plus",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("qwen-oauth"),
+            max_tokens_param_fn=_max_tokens_fn,
+        )
+        assert kw["max_completion_tokens"] == 65536
+
+    def test_vl_high_resolution(self, transport):
+        kw = transport.build_kwargs(
+            model="qwen3.5-plus",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("qwen-oauth"),
+        )
+        assert kw["extra_body"]["vl_high_resolution_images"] is True
+
+    def test_metadata_top_level(self, transport):
+        """Qwen metadata goes to top-level api_kwargs, NOT extra_body."""
+        meta = {"sessionId": "s123", "promptId": "p456"}
+        kw = transport.build_kwargs(
+            model="qwen3.5-plus",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("qwen-oauth"),
+            qwen_session_metadata=meta,
+        )
+        assert kw["metadata"] == meta
+        assert "metadata" not in kw.get("extra_body", {})
+
+
+class TestCustomOllamaParity:
+    """Custom/Ollama: num_ctx, think=false — now tested via profile."""
+
+    def test_ollama_num_ctx(self, transport):
+        kw = transport.build_kwargs(
+            model="llama3.1",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("custom"),
+            ollama_num_ctx=131072,
+        )
+        assert kw["extra_body"]["options"]["num_ctx"] == 131072
+
+    def test_think_false_when_disabled(self, transport):
+        kw = transport.build_kwargs(
+            model="qwen3:72b",
+            messages=_simple_messages(),
+            tools=None,
+            provider_profile=get_provider_profile("custom"),
+            reasoning_config={"enabled": False, "effort": "none"},
+        )
+        assert kw["extra_body"]["think"] is False

tests/run_agent/test_run_agent.py

@@ -1117,6 +1117,7 @@ class TestBuildApiKwargs:
         assert "temperature" not in kwargs
 
     def test_kimi_coding_endpoint_omits_temperature(self, agent):
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.kimi.com/coding/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-k2.5"
@@ -1129,6 +1130,7 @@ class TestBuildApiKwargs:
     def test_kimi_coding_endpoint_sends_max_tokens_and_reasoning(self, agent):
         """Kimi endpoint should send max_tokens=32000 and reasoning_effort as
         top-level params, matching Kimi CLI's default behavior."""
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.kimi.com/coding/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-for-coding"
@@ -1141,6 +1143,7 @@ class TestBuildApiKwargs:
 
     def test_kimi_coding_endpoint_respects_custom_effort(self, agent):
         """reasoning_effort should reflect reasoning_config.effort when set."""
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.kimi.com/coding/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-for-coding"
@@ -1154,6 +1157,7 @@ class TestBuildApiKwargs:
     def test_kimi_coding_endpoint_sends_thinking_extra_body(self, agent):
         """Kimi endpoint should send extra_body.thinking={"type":"enabled"}
         to activate reasoning mode, mirroring Kimi CLI's with_thinking()."""
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.kimi.com/coding/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-for-coding"
@@ -1167,6 +1171,7 @@ class TestBuildApiKwargs:
         """When reasoning_config.enabled=False, thinking should be disabled
         and reasoning_effort should be omitted entirely — mirroring Kimi
         CLI's with_thinking("off") which maps to reasoning_effort=None."""
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.kimi.com/coding/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-for-coding"
@@ -1180,6 +1185,7 @@ class TestBuildApiKwargs:
 
     def test_moonshot_endpoint_sends_max_tokens_and_reasoning(self, agent):
         """api.moonshot.ai should get the same Kimi-compatible params."""
+        agent.provider = "kimi-coding"
         agent.base_url = "https://api.moonshot.ai/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-k2.5"
@@ -1193,6 +1199,7 @@ class TestBuildApiKwargs:
 
     def test_moonshot_cn_endpoint_sends_max_tokens_and_reasoning(self, agent):
         """api.moonshot.cn (China endpoint) should get the same params."""
+        agent.provider = "kimi-coding-cn"
         agent.base_url = "https://api.moonshot.cn/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.model = "kimi-k2.5"
@@ -1205,6 +1212,7 @@ class TestBuildApiKwargs:
         assert kwargs["extra_body"]["thinking"] == {"type": "enabled"}
 
     def test_provider_preferences_injected(self, agent):
+        agent.provider = "openrouter"
         agent.base_url = "https://openrouter.ai/api/v1"
         agent.providers_allowed = ["Anthropic"]
         messages = [{"role": "user", "content": "hi"}]
@@ -1213,6 +1221,7 @@ class TestBuildApiKwargs:
 
     def test_reasoning_config_default_openrouter(self, agent):
         """Default reasoning config for OpenRouter should be medium."""
+        agent.provider = "openrouter"
         agent.base_url = "https://openrouter.ai/api/v1"
         agent.model = "anthropic/claude-sonnet-4-20250514"
         messages = [{"role": "user", "content": "hi"}]
@@ -1222,6 +1231,7 @@ class TestBuildApiKwargs:
         assert reasoning["effort"] == "medium"
 
     def test_reasoning_config_custom(self, agent):
+        agent.provider = "openrouter"
         agent.base_url = "https://openrouter.ai/api/v1"
         agent.model = "anthropic/claude-sonnet-4-20250514"
         agent.reasoning_config = {"enabled": False}
@@ -1237,6 +1247,7 @@ class TestBuildApiKwargs:
         assert "reasoning" not in kwargs.get("extra_body", {})
 
     def test_reasoning_sent_for_supported_openrouter_model(self, agent):
+        agent.provider = "openrouter"
         agent.base_url = "https://openrouter.ai/api/v1"
         agent.model = "qwen/qwen3.5-plus-02-15"
         messages = [{"role": "user", "content": "hi"}]
@@ -1244,6 +1255,7 @@ class TestBuildApiKwargs:
         assert kwargs["extra_body"]["reasoning"]["effort"] == "medium"
 
     def test_reasoning_sent_for_nous_route(self, agent):
+        agent.provider = "nous"
         agent.base_url = "https://inference-api.nousresearch.com/v1"
         agent.model = "minimax/minimax-m2.5"
         messages = [{"role": "user", "content": "hi"}]
@@ -1251,18 +1263,38 @@ class TestBuildApiKwargs:
         assert kwargs["extra_body"]["reasoning"]["effort"] == "medium"
 
     def test_reasoning_sent_for_copilot_gpt5(self, agent):
-        agent.base_url = "https://api.githubcopilot.com"
-        agent.model = "gpt-5.4"
-        messages = [{"role": "user", "content": "hi"}]
-        kwargs = agent._build_api_kwargs(messages)
+        """Copilot/GitHub Models: GPT-5 reasoning goes in extra_body.reasoning."""
+        from agent.transports import get_transport
+        from providers import get_provider_profile
+
+        transport = get_transport("chat_completions")
+        profile = get_provider_profile("copilot")
+        msgs = [{"role": "user", "content": "hi"}]
+        kwargs = transport.build_kwargs(
+            model="gpt-5.4",
+            messages=msgs,
+            tools=None,
+            supports_reasoning=True,
+            provider_profile=profile,
+        )
         assert kwargs["extra_body"]["reasoning"] == {"effort": "medium"}
 
     def test_reasoning_xhigh_normalized_for_copilot(self, agent):
-        agent.base_url = "https://api.githubcopilot.com"
-        agent.model = "gpt-5.4"
-        agent.reasoning_config = {"enabled": True, "effort": "xhigh"}
-        messages = [{"role": "user", "content": "hi"}]
-        kwargs = agent._build_api_kwargs(messages)
+        """xhigh effort should normalize to high for Copilot GitHub Models."""
+        from agent.transports import get_transport
+        from providers import get_provider_profile
+
+        transport = get_transport("chat_completions")
+        profile = get_provider_profile("copilot")
+        msgs = [{"role": "user", "content": "hi"}]
+        kwargs = transport.build_kwargs(
+            model="gpt-5.4",
+            messages=msgs,
+            tools=None,
+            supports_reasoning=True,
+            reasoning_config={"enabled": True, "effort": "xhigh"},
+            provider_profile=profile,
+        )
         assert kwargs["extra_body"]["reasoning"] == {"effort": "high"}
 
     def test_reasoning_omitted_for_non_reasoning_copilot_model(self, agent):
@@ -1280,6 +1312,7 @@ class TestBuildApiKwargs:
 
 
     def test_qwen_portal_formats_messages_and_metadata(self, agent):
+        agent.provider = "qwen-oauth"
         agent.base_url = "https://portal.qwen.ai/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.session_id = "sess-123"
@@ -1296,6 +1329,7 @@ class TestBuildApiKwargs:
         assert kwargs["messages"][2]["content"][0]["text"] == "hi"
 
     def test_qwen_portal_normalizes_bare_string_content_parts(self, agent):
+        agent.provider = "qwen-oauth"
         agent.base_url = "https://portal.qwen.ai/v1"
         agent._base_url_lower = agent.base_url.lower()
         messages = [
@@ -1308,6 +1342,7 @@ class TestBuildApiKwargs:
         assert user_content[1] == {"type": "text", "text": "world"}
 
     def test_qwen_portal_no_system_message(self, agent):
+        agent.provider = "qwen-oauth"
         agent.base_url = "https://portal.qwen.ai/v1"
         agent._base_url_lower = agent.base_url.lower()
         messages = [{"role": "user", "content": "hi"}]
@@ -1328,6 +1363,7 @@ class TestBuildApiKwargs:
     def test_qwen_portal_default_max_tokens(self, agent):
         """When max_tokens is None, Qwen Portal gets a default of 65536
         to prevent reasoning models from exhausting their output budget."""
+        agent.provider = "qwen-oauth"
         agent.base_url = "https://portal.qwen.ai/v1"
         agent._base_url_lower = agent.base_url.lower()
         agent.max_tokens = None

website/docs/developer-guide/adding-providers.md

@@ -93,6 +93,46 @@ This path includes everything from Path A plus:
 11. `run_agent.py`
 12. `pyproject.toml` if a provider SDK is required
 
+## Fast path: Simple API-key providers
+
+If your provider is just an OpenAI-compatible endpoint that authenticates with a single API key, you do not need to touch `auth.py`, `runtime_provider.py`, `main.py`, or any of the other files in the full checklist below.
+
+All you need is:
+
+1. A plugin directory under `plugins/model-providers/<your-provider>/` containing:
+   - `__init__.py` — calls `register_provider(profile)` at module-level
+   - `plugin.yaml` — manifest (name, kind: model-provider, version, description)
+2. That's it. Provider plugins auto-load the first time anything calls `get_provider_profile()` or `list_providers()` — bundled plugins (this repo) and user plugins at `$HERMES_HOME/plugins/model-providers/` both get picked up.
+
+When you add a plugin and it calls `register_provider()`, the following wire up automatically:
+
+1. `PROVIDER_REGISTRY` entry in `auth.py` (credential resolution, env-var lookup)
+2. `api_mode` set to `chat_completions`
+3. `base_url` sourced from the config or the declared env var
+4. `env_vars` checked in priority order for the API key
+5. `fallback_models` list registered for the provider
+6. `--provider` CLI flag accepts the provider id
+7. `hermes model` menu includes the provider
+8. `hermes setup` wizard delegates to `main.py` automatically
+9. `provider:model` alias syntax works
+10. Runtime resolver returns the correct `base_url` and `api_key`
+11. `HERMES_INFERENCE_PROVIDER` env-var override accepts the provider id
+12. Fallback model activation can switch into the provider cleanly
+
+User plugins at `$HERMES_HOME/plugins/model-providers/<name>/` override bundled plugins of the same name (last-writer-wins in `register_provider()`) — so third parties can monkey-patch or replace any built-in profile without editing the repo.
+
+See `plugins/model-providers/nvidia/` or `plugins/model-providers/gmi/` as a template, and the full [Model Provider Plugin guide](/docs/developer-guide/model-provider-plugin) for field reference, hook idioms, and end-to-end examples.
+
+## Full path: OAuth and complex providers
+
+Use the full checklist below when your provider needs any of the following:
+
+- OAuth or token refresh (Nous Portal, Codex, Google Gemini, Qwen Portal, Copilot)
+- A non-OpenAI API shape that requires a new adapter (Anthropic Messages, Codex Responses)
+- Custom endpoint detection or multi-region probing (z.ai, Kimi)
+- A curated static model catalog or live `/models` fetch
+- Provider-specific `hermes model` menu entries with bespoke auth flows
+
 ## Step 1: Pick one canonical provider id
 
 Choose a single provider id and use it everywhere.

website/docs/developer-guide/model-provider-plugin.md

@@ -0,0 +1,267 @@
+---
+sidebar_position: 10
+title: "Model Provider Plugins"
+description: "How to build a model provider (inference backend) plugin for Hermes Agent"
+---
+
+# Building a Model Provider Plugin
+
+Model provider plugins declare an inference backend — an OpenAI-compatible endpoint, an Anthropic Messages server, a Codex-style Responses API, or a Bedrock-native surface — that Hermes can route `AIAgent` calls through. Every built-in provider (OpenRouter, Anthropic, GMI, DeepSeek, Nvidia, …) ships as one of these plugins. Third parties can add their own by dropping a directory under `$HERMES_HOME/plugins/model-providers/` with zero changes to the repo.
+
+:::tip
+Model provider plugins are the third kind of **provider plugin**. The others are [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) (cross-session knowledge) and [Context Engine Plugins](/docs/developer-guide/context-engine-plugin) (context compression strategies). All three follow the same "drop a directory, declare a profile, no repo edits" pattern.
+:::
+
+## How discovery works
+
+`providers/__init__.py._discover_providers()` runs lazily the first time any code calls `get_provider_profile()` or `list_providers()`. Discovery order:
+
+1. **Bundled plugins** — `<repo>/plugins/model-providers/<name>/` — ship with Hermes
+2. **User plugins** — `$HERMES_HOME/plugins/model-providers/<name>/` — drop in any directory; no restart required for subsequent sessions
+3. **Legacy single-file** — `<repo>/providers/<name>.py` — back-compat for out-of-tree editable installs
+
+**User plugins override bundled plugins of the same name** because `register_provider()` is last-writer-wins. Drop a `$HERMES_HOME/plugins/model-providers/gmi/` directory to replace the built-in GMI profile without touching the repo.
+
+## Directory structure
+
+```
+plugins/model-providers/my-provider/
+├── __init__.py       # Calls register_provider(profile) at module-level
+├── plugin.yaml       # kind: model-provider + metadata (optional but recommended)
+└── README.md         # Setup instructions (optional)
+```
+
+The only required file is `__init__.py`. `plugin.yaml` is used by `hermes plugins` for introspection and by the general PluginManager to route the plugin to the right loader; without it, the general loader falls back to a source-text heuristic.
+
+## Minimal example — a simple API-key provider
+
+```python
+# plugins/model-providers/acme-inference/__init__.py
+from providers import register_provider
+from providers.base import ProviderProfile
+
+acme = ProviderProfile(
+    name="acme-inference",
+    aliases=("acme",),
+    display_name="Acme Inference",
+    description="Acme — OpenAI-compatible direct API",
+    signup_url="https://acme.example.com/keys",
+    env_vars=("ACME_API_KEY", "ACME_BASE_URL"),
+    base_url="https://api.acme.example.com/v1",
+    auth_type="api_key",
+    default_aux_model="acme-small-fast",
+    fallback_models=(
+        "acme-large-v3",
+        "acme-medium-v3",
+        "acme-small-fast",
+    ),
+)
+
+register_provider(acme)
+```
+
+```yaml
+# plugins/model-providers/acme-inference/plugin.yaml
+name: acme-inference
+kind: model-provider
+version: 1.0.0
+description: Acme Inference — OpenAI-compatible direct API
+author: Your Name
+```
+
+That's it. After dropping these two files, the following **auto-wire** with no other edits:
+
+| Integration | Where | What it gets |
+|---|---|---|
+| Credential resolution | `hermes_cli/auth.py` | `PROVIDER_REGISTRY["acme-inference"]` populated from profile |
+| `--provider` CLI flag | `hermes_cli/main.py` | Accepts `acme-inference` |
+| `hermes model` picker | `hermes_cli/models.py` | Appears in `CANONICAL_PROVIDERS`, model list fetched from `{base_url}/models` |
+| `hermes doctor` | `hermes_cli/doctor.py` | Health check for `ACME_API_KEY` + `{base_url}/models` probe |
+| `hermes setup` | `hermes_cli/config.py` | `ACME_API_KEY` appears in `OPTIONAL_ENV_VARS` and the setup wizard |
+| URL reverse-mapping | `agent/model_metadata.py` | Hostname → provider name for auto-detection |
+| Auxiliary model | `agent/auxiliary_client.py` | Uses `default_aux_model` for compression / summarization |
+| Runtime resolution | `hermes_cli/runtime_provider.py` | Returns correct `base_url`, `api_key`, `api_mode` |
+| Transport | `agent/transports/chat_completions.py` | Profile path generates kwargs via `prepare_messages` / `build_extra_body` / `build_api_kwargs_extras` |
+
+## ProviderProfile fields
+
+Full definition in `providers/base.py`. The most useful ones:
+
+| Field | Type | Purpose |
+|---|---|---|
+| `name` | str | Canonical id — matches `--provider` choices and `HERMES_INFERENCE_PROVIDER` |
+| `aliases` | `tuple[str, ...]` | Alternative names resolved by `get_provider_profile()` (e.g. `grok` → `xai`) |
+| `api_mode` | str | `chat_completions` \| `codex_responses` \| `anthropic_messages` \| `bedrock_converse` |
+| `display_name` | str | Human label shown in `hermes model` picker |
+| `description` | str | Picker subtitle |
+| `signup_url` | str | Shown during first-run setup ("get an API key here") |
+| `env_vars` | `tuple[str, ...]` | API-key env vars in priority order; a final `*_BASE_URL` entry is used as the user base-URL override |
+| `base_url` | str | Default inference endpoint |
+| `models_url` | str | Explicit catalog URL (falls back to `{base_url}/models`) |
+| `auth_type` | str | `api_key` \| `oauth_device_code` \| `oauth_external` \| `copilot` \| `aws_sdk` \| `external_process` |
+| `fallback_models` | `tuple[str, ...]` | Curated list shown when live catalog fetch fails |
+| `default_headers` | `dict[str, str]` | Sent on every request (e.g. Copilot's `Editor-Version`) |
+| `fixed_temperature` | Any | `None` = use caller's value; `OMIT_TEMPERATURE` sentinel = don't send temperature at all (Kimi) |
+| `default_max_tokens` | `int \| None` | Provider-level max_tokens cap (Nvidia: 16384) |
+| `default_aux_model` | str | Cheap model for auxiliary tasks (compression, vision, summarization) |
+
+## Overridable hooks
+
+Subclass `ProviderProfile` for non-trivial quirks:
+
+```python
+from typing import Any
+from providers.base import ProviderProfile
+
+class AcmeProfile(ProviderProfile):
+    def prepare_messages(self, messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Provider-specific message preprocessing. Runs after codex
+        sanitization, before developer-role swap. Default: pass-through."""
+        # Example: Qwen normalizes plain-text content to a list-of-parts
+        # array and injects cache_control; Kimi rewrites tool-call JSON
+        return messages
+
+    def build_extra_body(self, *, session_id=None, **context) -> dict:
+        """Provider-specific extra_body fields merged into the API call.
+        Context includes: session_id, provider_preferences, model, base_url,
+        reasoning_config. Default: empty dict."""
+        # Example: OpenRouter's provider-preferences block,
+        # Gemini's thinking_config translation.
+        return {}
+
+    def build_api_kwargs_extras(self, *, reasoning_config=None, **context):
+        """Returns (extra_body_additions, top_level_kwargs). Needed when some
+        fields go top-level (Kimi's reasoning_effort) and some go in extra_body
+        (OpenRouter's reasoning dict). Default: ({}, {})."""
+        return {}, {}
+
+    def fetch_models(self, *, api_key=None, timeout=8.0) -> list[str] | None:
+        """Live catalog fetch. Default hits {models_url or base_url}/models with
+        Bearer auth. Override for: custom auth (Anthropic), no REST endpoint
+        (Bedrock → None), or public/unauthenticated catalogs (OpenRouter)."""
+        return super().fetch_models(api_key=api_key, timeout=timeout)
+```
+
+## Hook reference examples
+
+Look at these bundled plugins for idioms:
+
+| Plugin | Why look |
+|---|---|
+| `plugins/model-providers/openrouter/` | Aggregator with provider preferences, public model catalog |
+| `plugins/model-providers/gemini/` | `thinking_config` translation (native + OpenAI-compat nested forms) |
+| `plugins/model-providers/kimi-coding/` | `OMIT_TEMPERATURE`, `extra_body.thinking`, top-level `reasoning_effort` |
+| `plugins/model-providers/qwen-oauth/` | Message normalization, `cache_control` injection, VL high-res |
+| `plugins/model-providers/nous/` | Attribution tags, "omit reasoning when disabled" |
+| `plugins/model-providers/custom/` | Ollama `num_ctx` + `think: false` quirks |
+| `plugins/model-providers/bedrock/` | `api_mode="bedrock_converse"`, `fetch_models` returns None (no REST endpoint) |
+
+## User overrides — replace a built-in without editing the repo
+
+Say you want to point `gmi` at your private staging endpoint for testing. Create `~/.hermes/plugins/model-providers/gmi/__init__.py`:
+
+```python
+from providers import register_provider
+from providers.base import ProviderProfile
+
+register_provider(ProviderProfile(
+    name="gmi",
+    aliases=("gmi-cloud", "gmicloud"),
+    env_vars=("GMI_API_KEY",),
+    base_url="https://gmi-staging.internal.example.com/v1",
+    auth_type="api_key",
+    default_aux_model="google/gemini-3.1-flash-lite-preview",
+))
+```
+
+Next session, `get_provider_profile("gmi").base_url` returns the staging URL. No repo patch, no rebuild. Because user plugins are discovered after bundled ones, the user `register_provider()` call wins.
+
+## api_mode selection
+
+Four values are recognized. Hermes picks one based on:
+
+1. User explicit override (`config.yaml` `model.api_mode` when set)
+2. OpenCode's per-model dispatch (`opencode_model_api_mode` for Zen and Go)
+3. URL auto-detection — `/anthropic` suffix → `anthropic_messages`, `api.openai.com` → `codex_responses`, `api.x.ai` → `codex_responses`, `/coding` on Kimi domains → `chat_completions`
+4. **Profile `api_mode`** as a fallback when URL detection finds nothing
+5. Default `chat_completions`
+
+Set `profile.api_mode` to match the default your provider ships — it acts as a hint. User URL overrides still win.
+
+## Auth types
+
+| `auth_type` | Meaning | Who uses it |
+|---|---|---|
+| `api_key` | Single env var carries a static API key | Most providers |
+| `oauth_device_code` | Device-code OAuth flow | — |
+| `oauth_external` | User signs in elsewhere, tokens land in `auth.json` | Anthropic OAuth, MiniMax OAuth, Gemini Cloud Code, Qwen Portal, Nous Portal |
+| `copilot` | GitHub Copilot token refresh cycle | `copilot` plugin only |
+| `aws_sdk` | AWS SDK credential chain (IAM role, profile, env) | `bedrock` plugin only |
+| `external_process` | Auth handled by a subprocess the agent spawns | `copilot-acp` plugin only |
+
+`auth_type` gates which codepaths treat your provider as a "simple api-key provider" — if it's not `api_key`, the PluginManager still records the manifest but Hermes' CLI-level automation (doctor checks, `--provider` flag, setup wizard delegation) may skip over it.
+
+## Discovery timing
+
+Provider discovery is **lazy** — triggered by the first `get_provider_profile()` or `list_providers()` call in the process. In practice this happens early at startup (`auth.py` module load extends `PROVIDER_REGISTRY` eagerly). If you need to verify your plugin loaded, run:
+
+```bash
+hermes doctor
+```
+
+— a successful `auth_type="api_key"` profile appears under the Provider Connectivity section with a `/models` probe.
+
+For programmatic inspection:
+
+```python
+from providers import list_providers
+for p in list_providers():
+    print(p.name, p.base_url, p.api_mode)
+```
+
+## Testing your plugin
+
+Point `HERMES_HOME` at a temp directory so you don't pollute your real config:
+
+```bash
+export HERMES_HOME=/tmp/hermes-plugin-test
+mkdir -p $HERMES_HOME/plugins/model-providers/my-provider
+cat > $HERMES_HOME/plugins/model-providers/my-provider/__init__.py <<'EOF'
+from providers import register_provider
+from providers.base import ProviderProfile
+register_provider(ProviderProfile(
+    name="my-provider",
+    env_vars=("MY_API_KEY",),
+    base_url="https://api.my-provider.example.com/v1",
+    auth_type="api_key",
+))
+EOF
+
+export MY_API_KEY=your-test-key
+hermes -z "hello" --provider my-provider -m some-model
+```
+
+## General PluginManager integration
+
+The general `PluginManager` (the thing `hermes plugins` operates on) **sees** model-provider plugins but does not import them — `providers/__init__.py` owns their lifecycle. The manager records the manifest for introspection and categorizes by `kind: model-provider`. When you drop an unlabeled user plugin into `$HERMES_HOME/plugins/` that happens to call `register_provider` with a `ProviderProfile`, the manager auto-coerces it to `kind: model-provider` via a source-text heuristic — so the plugin still routes correctly even without `plugin.yaml`.
+
+## Distribute via pip
+
+Like any Hermes plugin, model providers can ship as a pip package. Add an entry point to your `pyproject.toml`:
+
+```toml
+[project.entry-points."hermes.plugins"]
+acme-inference = "acme_hermes_plugin:register"
+```
+
+…where `acme_hermes_plugin:register` is a function that calls `register_provider(profile)`. The general PluginManager picks up entry-point plugins during `discover_and_load()`. For `kind: model-provider` pip plugins, you still need to declare the kind in your manifest (or rely on the source-text heuristic).
+
+See [Building a Hermes Plugin](/docs/guides/build-a-hermes-plugin#distribute-via-pip) for the full entry-points setup.
+
+## Related pages
+
+- [Provider Runtime](/docs/developer-guide/provider-runtime) — resolution precedence + where each layer reads the profile
+- [Adding Providers](/docs/developer-guide/adding-providers) — end-to-end checklist for new inference backends (covers both the fast plugin path and the full CLI/auth integration)
+- [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin)
+- [Context Engine Plugins](/docs/developer-guide/context-engine-plugin)
+- [Building a Hermes Plugin](/docs/guides/build-a-hermes-plugin) — general plugin authoring

website/docs/developer-guide/provider-runtime.md

@@ -20,8 +20,12 @@ Primary implementation:
 - `hermes_cli/auth.py` — provider registry, `resolve_provider()`
 - `hermes_cli/model_switch.py` — shared `/model` switch pipeline (CLI + gateway)
 - `agent/auxiliary_client.py` — auxiliary model routing
+- `providers/` — ABC + registry entry points (`ProviderProfile`, `register_provider`, `get_provider_profile`, `list_providers`)
+- `plugins/model-providers/<name>/` — per-provider plugins (bundled) that declare `api_mode`, `base_url`, `env_vars`, `fallback_models` and register themselves into the registry on first access. User plugins at `$HERMES_HOME/plugins/model-providers/<name>/` override bundled ones of the same name.
 
-If you are trying to add a new first-class inference provider, read [Adding Providers](./adding-providers.md) alongside this page.
+`get_provider_profile()` in `providers/` returns a `ProviderProfile` for a given provider id. `runtime_provider.py` calls this at resolution time to get the canonical `base_url`, `env_vars` priority list, `api_mode`, and `fallback_models` without needing to duplicate that data in multiple files. Adding a new plugin under `plugins/model-providers/<your-provider>/` (or `$HERMES_HOME/plugins/model-providers/<your-provider>/`) that calls `register_provider()` is enough for `runtime_provider.py` to pick it up — no branch needed in the resolver itself.
+
+If you are trying to add a new first-class inference provider, read [Adding Providers](./adding-providers.md) and the [Model Provider Plugin guide](./model-provider-plugin.md) alongside this page.
 
 ## Resolution precedence
 

website/docs/guides/build-a-hermes-plugin.md

@@ -9,6 +9,28 @@ description: "Step-by-step guide to building a complete Hermes plugin with tools
 
 This guide walks through building a complete Hermes plugin from scratch. By the end you'll have a working plugin with multiple tools, lifecycle hooks, shipped data files, and a bundled skill — everything the plugin system supports.
 
+:::info Not sure which guide you need?
+Hermes has several distinct pluggable interfaces — some use Python `register_*` APIs, others are config-driven or drop-in directories. Use this map first:
+
+| If you want to add… | Read |
+|---|---|
+| Custom tools, hooks, slash commands, skills, or CLI subcommands | **This guide** (the general plugin surface) |
+| An **LLM / inference backend** (new provider) | [Model Provider Plugins](/docs/developer-guide/model-provider-plugin) |
+| A **gateway channel** (Discord/Telegram/IRC/Teams/etc.) | [Adding Platform Adapters](/docs/developer-guide/adding-platform-adapters) |
+| A **memory backend** (Honcho/Mem0/Supermemory/etc.) | [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) |
+| A **context-compression engine** | [Context Engine Plugins](/docs/developer-guide/context-engine-plugin) |
+| An **image-generation backend** | See bundled examples in `plugins/image_gen/openai/` and `plugins/image_gen/xai/` |
+| A **TTS backend** (any CLI — Piper, VoxCPM, Kokoro, voice cloning, …) | [TTS custom command providers](/docs/user-guide/features/tts#custom-command-providers) — config-driven, no Python needed |
+| An **STT backend** (custom whisper / ASR CLI) | [Voice Message Transcription](/docs/user-guide/features/tts#voice-message-transcription-stt) — set `HERMES_LOCAL_STT_COMMAND` to a shell template |
+| **External tools via MCP** (filesystem, GitHub, Linear, any MCP server) | [MCP](/docs/user-guide/features/mcp) — declare `mcp_servers.<name>` in `config.yaml` |
+| **Gateway event hooks** (fire on startup, session events, commands) | [Event Hooks](/docs/user-guide/features/hooks#gateway-event-hooks) — drop `HOOK.yaml` + `handler.py` into `~/.hermes/hooks/<name>/` |
+| **Shell hooks** (run a shell command on events) | [Shell Hooks](/docs/user-guide/features/hooks#shell-hooks) — declare under `hooks:` in `config.yaml` |
+| **Additional skill sources** (custom GitHub repos, private skill indexes) | [Skills](/docs/user-guide/features/skills) — `hermes skills tap add <repo>` |
+| A first-class **core** inference provider (not a plugin) | [Adding Providers](/docs/developer-guide/adding-providers) |
+
+See the full [Pluggable interfaces table](/docs/user-guide/features/plugins#pluggable-interfaces--where-to-go-for-each) for a consolidated view of every extension surface including config-driven (TTS, STT, MCP, shell hooks) and drop-in directory (gateway hooks) styles.
+:::
+
 ## What you're building
 
 A **calculator** plugin with two tools:
@@ -629,12 +651,267 @@ def register(ctx):
 ```
 
 :::tip
-This guide covers **general plugins** (tools, hooks, slash commands, CLI commands). For specialized plugin types, see:
-- [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) — cross-session knowledge backends
-- [Context Engine Plugins](/docs/developer-guide/context-engine-plugin) — alternative context management strategies
+This guide covers **general plugins** (tools, hooks, slash commands, CLI commands). The sections below sketch the authoring pattern for each specialized plugin type; each links to its full guide for field reference and examples.
 :::
 
-### Distribute via pip
+## Specialized plugin types
+
+Hermes has five specialized plugin types beyond the general surface. Each ships as a directory under `plugins/<category>/<name>/` (bundled) or `~/.hermes/plugins/<category>/<name>/` (user). The contract differs by category — pick the one you need, then read its full guide.
+
+### Model provider plugins — add an LLM backend
+
+Drop a profile into `plugins/model-providers/<name>/`:
+
+```python
+# plugins/model-providers/acme/__init__.py
+from providers import register_provider
+from providers.base import ProviderProfile
+
+register_provider(ProviderProfile(
+    name="acme",
+    aliases=("acme-inference",),
+    display_name="Acme Inference",
+    env_vars=("ACME_API_KEY", "ACME_BASE_URL"),
+    base_url="https://api.acme.example.com/v1",
+    auth_type="api_key",
+    default_aux_model="acme-small-fast",
+    fallback_models=("acme-large-v3", "acme-medium-v3"),
+))
+```
+
+```yaml
+# plugins/model-providers/acme/plugin.yaml
+name: acme-provider
+kind: model-provider
+version: 1.0.0
+description: Acme Inference — OpenAI-compatible direct API
+```
+
+Lazy-discovered the first time anything calls `get_provider_profile()` or `list_providers()` — `auth.py`, `config.py`, `doctor.py`, `models.py`, `runtime_provider.py`, and the chat_completions transport auto-wire to it. User plugins override bundled ones by name.
+
+**Full guide:** [Model Provider Plugins](/docs/developer-guide/model-provider-plugin) — field reference, overridable hooks (`prepare_messages`, `build_extra_body`, `build_api_kwargs_extras`, `fetch_models`), api_mode selection, auth types, testing.
+
+### Platform plugins — add a gateway channel
+
+Drop an adapter into `plugins/platforms/<name>/`:
+
+```python
+# plugins/platforms/myplatform/adapter.py
+from gateway.platforms.base import BasePlatformAdapter
+
+class MyPlatformAdapter(BasePlatformAdapter):
+    async def connect(self): ...
+    async def send(self, chat_id, text): ...
+    async def disconnect(self): ...
+
+def check_requirements():
+    import os
+    return bool(os.environ.get("MYPLATFORM_TOKEN"))
+
+def register(ctx):
+    ctx.register_platform(
+        name="myplatform",
+        label="MyPlatform",
+        adapter_factory=lambda cfg: MyPlatformAdapter(cfg),
+        check_fn=check_requirements,
+        required_env=["MYPLATFORM_TOKEN"],
+        emoji="💬",
+        platform_hint="You are chatting via MyPlatform. Keep responses concise.",
+    )
+```
+
+```yaml
+# plugins/platforms/myplatform/plugin.yaml
+name: myplatform-platform
+kind: platform
+version: 1.0.0
+description: MyPlatform gateway adapter
+requires_env: [MYPLATFORM_TOKEN]
+```
+
+**Full guide:** [Adding Platform Adapters](/docs/developer-guide/adding-platform-adapters) — complete `BasePlatformAdapter` contract, message routing, auth gating, setup wizard integration. Look at `plugins/platforms/irc/` for a stdlib-only working example.
+
+### Memory provider plugins — add a cross-session knowledge backend
+
+Drop an implementation of `MemoryProvider` into `plugins/memory/<name>/`:
+
+```python
+# plugins/memory/my-memory/__init__.py
+from agent.memory_provider import MemoryProvider
+
+class MyMemoryProvider(MemoryProvider):
+    @property
+    def name(self) -> str:
+        return "my-memory"
+
+    def is_available(self) -> bool:
+        import os
+        return bool(os.environ.get("MY_MEMORY_API_KEY"))
+
+    def initialize(self, session_id: str, **kwargs) -> None:
+        self._session_id = session_id
+
+    def sync_turn(self, user_message, assistant_response, **kwargs) -> None:
+        ...
+
+    def prefetch(self, query: str, **kwargs) -> str | None:
+        ...
+
+def register(ctx):
+    ctx.register_memory_provider(MyMemoryProvider())
+```
+
+Memory providers are single-select — only one is active at a time, chosen via `memory.provider` in `config.yaml`.
+
+**Full guide:** [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) — full `MemoryProvider` ABC, threading contract, profile isolation, CLI command registration via `cli.py`.
+
+### Context engine plugins — replace the context compressor
+
+```python
+# plugins/context_engine/my-engine/__init__.py
+from agent.context_engine import ContextEngine
+
+class MyContextEngine(ContextEngine):
+    @property
+    def name(self) -> str:
+        return "my-engine"
+
+    def should_compress(self, messages, model) -> bool: ...
+    def compress(self, messages, model) -> list[dict]: ...
+
+def register(ctx):
+    ctx.register_context_engine(MyContextEngine())
+```
+
+Context engines are single-select — chosen via `context.engine` in `config.yaml`.
+
+**Full guide:** [Context Engine Plugins](/docs/developer-guide/context-engine-plugin).
+
+### Image-generation backends
+
+Drop a provider into `plugins/image_gen/<name>/`:
+
+```python
+# plugins/image_gen/my-imggen/__init__.py
+from agent.image_gen_provider import ImageGenProvider
+
+class MyImageGenProvider(ImageGenProvider):
+    @property
+    def name(self) -> str:
+        return "my-imggen"
+
+    def is_available(self) -> bool: ...
+    def generate(self, prompt: str, **kwargs) -> str: ...   # returns image path
+
+def register(ctx):
+    ctx.register_image_gen_provider(MyImageGenProvider())
+```
+
+```yaml
+# plugins/image_gen/my-imggen/plugin.yaml
+name: my-imggen
+kind: backend
+version: 1.0.0
+description: Custom image generation backend
+```
+
+**Reference examples:** `plugins/image_gen/openai/` (DALL-E / GPT-Image via OpenAI SDK), `plugins/image_gen/openai-codex/`, `plugins/image_gen/xai/` (Grok image gen).
+
+## Non-Python extension surfaces
+
+Hermes also accepts extensions that aren't Python plugins at all. These are shown in the [Pluggable interfaces table](/docs/user-guide/features/plugins#pluggable-interfaces--where-to-go-for-each); the sections below sketch each authoring style briefly.
+
+### MCP servers — register external tools
+
+Model Context Protocol (MCP) servers register their own tools into Hermes without any Python plugin. Declare them in `~/.hermes/config.yaml`:
+
+```yaml
+mcp_servers:
+  filesystem:
+    command: "npx"
+    args: ["-y", "@modelcontextprotocol/server-filesystem", "/home/user/projects"]
+    timeout: 120
+
+  linear:
+    url: "https://mcp.linear.app/sse"
+    auth:
+      type: "oauth"
+```
+
+Hermes connects to each server at startup, lists its tools, and registers them alongside built-ins. The LLM sees them exactly like any other tool. **Full guide:** [MCP](/docs/user-guide/features/mcp).
+
+### Gateway event hooks — fire on lifecycle events
+
+Drop a manifest + handler into `~/.hermes/hooks/<name>/`:
+
+```yaml
+# ~/.hermes/hooks/long-task-alert/HOOK.yaml
+name: long-task-alert
+description: Send a push notification when a long task finishes
+events:
+  - agent:end
+```
+
+```python
+# ~/.hermes/hooks/long-task-alert/handler.py
+async def handle(event_type: str, context: dict) -> None:
+    if context.get("duration_seconds", 0) > 120:
+        # send notification …
+        pass
+```
+
+Events include `gateway:startup`, `session:start`, `session:end`, `session:reset`, `agent:start`, `agent:step`, `agent:end`, and wildcard `command:*`. Errors in hooks are caught and logged — they never block the main pipeline.
+
+**Full guide:** [Gateway Event Hooks](/docs/user-guide/features/hooks#gateway-event-hooks).
+
+### Shell hooks — run a shell command on tool calls
+
+If you just want to run a script when a tool fires (notifications, audit logs, desktop alerts, auto-formatters), use shell hooks in `config.yaml` — no Python required:
+
+```yaml
+hooks:
+  - event: post_tool_call
+    command: "notify-send 'Tool ran: {tool_name}'"
+    when:
+      tools: [terminal, patch, write_file]
+```
+
+Supports all the same events as Python plugin hooks (`pre_tool_call`, `post_tool_call`, `pre_llm_call`, `post_llm_call`, `on_session_start`, `on_session_end`, `pre_gateway_dispatch`) plus structured JSON output for `pre_tool_call` blocking decisions.
+
+**Full guide:** [Shell Hooks](/docs/user-guide/features/hooks#shell-hooks).
+
+### Skill sources — add a custom skill registry
+
+If you maintain a private GitHub repo of skills (or want to pull from a community index beyond the built-in sources), add it as a **tap**:
+
+```bash
+hermes skills tap add myorg/skills-repo
+hermes skills search my-workflow --source myorg/skills-repo
+hermes skills install myorg/skills-repo/my-workflow
+```
+
+**Full guide:** [Skills Hub](/docs/user-guide/features/skills#skills-hub).
+
+### TTS / STT via command templates
+
+Any CLI that reads/writes audio or text can be plugged in through `config.yaml` — no Python code:
+
+```yaml
+tts:
+  provider: voxcpm
+  providers:
+    voxcpm:
+      type: command
+      command: "voxcpm --ref ~/voice.wav --text-file {input_path} --out {output_path}"
+      output_format: mp3
+      voice_compatible: true
+```
+
+For STT, point `HERMES_LOCAL_STT_COMMAND` at a shell template. Supported placeholders: `{input_path}`, `{output_path}`, `{format}`, `{voice}`, `{model}`, `{speed}` (TTS); `{input_path}`, `{output_dir}`, `{language}`, `{model}` (STT). Any path-interacting CLI is automatically a plugin.
+
+**Full guides:** [TTS custom command providers](/docs/user-guide/features/tts#custom-command-providers) · [STT](/docs/user-guide/features/tts#voice-message-transcription-stt).
+
+## Distribute via pip
 
 For sharing plugins publicly, add an entry point to your Python package:
 
@@ -649,7 +926,7 @@ pip install hermes-plugin-calculator
 # Plugin auto-discovered on next hermes startup
 ```
 
-### Distribute for NixOS
+## Distribute for NixOS
 
 NixOS users can install your plugin declaratively if you provide a `pyproject.toml` with entry points:
 

website/docs/integrations/providers.md

@@ -480,6 +480,44 @@ model:
 For on-prem deployments (DGX Spark, local GPU), set `NVIDIA_BASE_URL=http://localhost:8000/v1`. NIM exposes the same OpenAI-compatible chat completions API as build.nvidia.com, so switching between cloud and local is a one-line env-var change.
 :::
 
+### GMI Cloud
+
+Open and reasoning models via [GMI Cloud](https://inference.gmi.ai) — OpenAI-compatible API, API key authentication.
+
+```bash
+# GMI Cloud
+hermes chat --provider gmi --model deepseek-ai/DeepSeek-R1
+# Requires: GMI_API_KEY in ~/.hermes/.env
+```
+
+Or set it permanently in `config.yaml`:
+```yaml
+model:
+  provider: "gmi"
+  default: "deepseek-ai/DeepSeek-R1"
+```
+
+The base URL can be overridden with `GMI_BASE_URL` (default: `https://api.gmi.ai/v1`).
+
+### StepFun
+
+Step-series models via [StepFun](https://platform.stepfun.com) — OpenAI-compatible API, API key authentication.
+
+```bash
+# StepFun
+hermes chat --provider stepfun --model step-3-mini
+# Requires: STEPFUN_API_KEY in ~/.hermes/.env
+```
+
+Or set it permanently in `config.yaml`:
+```yaml
+model:
+  provider: "stepfun"
+  default: "step-3-mini"
+```
+
+The base URL can be overridden with `STEPFUN_BASE_URL` (default: `https://api.stepfun.com/v1`).
+
 ### Hugging Face Inference Providers
 
 [Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers) routes to 20+ open models through a unified OpenAI-compatible endpoint (`router.huggingface.co/v1`). Requests are automatically routed to the fastest available backend (Groq, Together, SambaNova, etc.) with automatic failover.
@@ -1239,7 +1277,7 @@ fallback_model:
 
 When activated, the fallback swaps the model and provider mid-session without losing your conversation. It fires **at most once** per session.
 
-Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `huggingface`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `ollama-cloud`, `bedrock`, `ai-gateway`, `opencode-zen`, `opencode-go`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `alibaba`, `tencent-tokenhub`, `custom`.
+Supported providers: `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `gemini`, `google-gemini-cli`, `qwen-oauth`, `huggingface`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth`, `deepseek`, `nvidia`, `xai`, `ollama-cloud`, `bedrock`, `ai-gateway`, `opencode-zen`, `opencode-go`, `kilocode`, `xiaomi`, `arcee`, `gmi`, `stepfun`, `alibaba`, `tencent-tokenhub`, `custom`.
 
 :::tip
 Fallback is configured exclusively through `config.yaml` — there are no environment variables for it. For full details on when it triggers, supported providers, and how it interacts with auxiliary tasks and delegation, see [Fallback Providers](/docs/user-guide/features/fallback-providers).

website/docs/reference/environment-variables.md

@@ -69,6 +69,10 @@ All variables go in `~/.hermes/.env`. You can also set them with `hermes config
 | `DEEPSEEK_BASE_URL` | Custom DeepSeek API base URL |
 | `NVIDIA_API_KEY` | NVIDIA NIM API key — Nemotron and open models ([build.nvidia.com](https://build.nvidia.com)) |
 | `NVIDIA_BASE_URL` | Override NVIDIA base URL (default: `https://integrate.api.nvidia.com/v1`; set to `http://localhost:8000/v1` for a local NIM endpoint) |
+| `GMI_API_KEY` | GMI Cloud API key — open and reasoning models ([inference.gmi.ai](https://inference.gmi.ai)) |
+| `GMI_BASE_URL` | Override GMI Cloud base URL (default: `https://api.gmi.ai/v1`) |
+| `STEPFUN_API_KEY` | StepFun API key — Step-series models ([platform.stepfun.com](https://platform.stepfun.com)) |
+| `STEPFUN_BASE_URL` | Override StepFun base URL (default: `https://api.stepfun.com/v1`) |
 | `OLLAMA_API_KEY` | Ollama Cloud API key — managed Ollama catalog without local GPU ([ollama.com/settings/keys](https://ollama.com/settings/keys)) |
 | `OLLAMA_BASE_URL` | Override Ollama Cloud base URL (default: `https://ollama.com/v1`) |
 | `XAI_API_KEY` | xAI (Grok) API key for chat + TTS ([console.x.ai](https://console.x.ai/)) |
@@ -99,7 +103,7 @@ For native Anthropic auth, Hermes prefers Claude Code's own credential files whe
 
 | Variable | Description |
 |----------|-------------|
-| `HERMES_INFERENCE_PROVIDER` | Override provider selection: `auto`, `custom`, `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `huggingface`, `gemini`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth` (browser OAuth login — no API key required; see [MiniMax OAuth guide](../guides/minimax-oauth.md)), `kilocode`, `xiaomi`, `arcee`, `gmi`, `alibaba`, `alibaba-coding-plan` (alias `alibaba_coding`), `deepseek`, `nvidia`, `ollama-cloud`, `xai` (alias `grok`), `google-gemini-cli`, `qwen-oauth`, `bedrock`, `opencode-zen`, `opencode-go`, `ai-gateway`, `tencent-tokenhub` (default: `auto`) |
+| `HERMES_INFERENCE_PROVIDER` | Override provider selection: `auto`, `custom`, `openrouter`, `nous`, `openai-codex`, `copilot`, `copilot-acp`, `anthropic`, `huggingface`, `gemini`, `zai`, `kimi-coding`, `kimi-coding-cn`, `minimax`, `minimax-cn`, `minimax-oauth` (browser OAuth login — no API key required; see [MiniMax OAuth guide](../guides/minimax-oauth.md)), `kilocode`, `xiaomi`, `arcee`, `gmi`, `stepfun`, `alibaba`, `alibaba-coding-plan` (alias `alibaba_coding`), `deepseek`, `nvidia`, `ollama-cloud`, `xai` (alias `grok`), `google-gemini-cli`, `qwen-oauth`, `bedrock`, `opencode-zen`, `opencode-go`, `ai-gateway`, `tencent-tokenhub` (default: `auto`) |
 | `HERMES_PORTAL_BASE_URL` | Override Nous Portal URL (for development/testing) |
 | `NOUS_INFERENCE_BASE_URL` | Override Nous inference API URL |
 | `HERMES_NOUS_MIN_KEY_TTL_SECONDS` | Min agent key TTL before re-mint (default: 1800 = 30min) |

website/docs/user-guide/features/fallback-providers.md

@@ -60,6 +60,8 @@ Both `provider` and `model` are **required**. If either is missing, the fallback
 | MiniMax (China) | `minimax-cn` | `MINIMAX_CN_API_KEY` |
 | DeepSeek | `deepseek` | `DEEPSEEK_API_KEY` |
 | NVIDIA NIM | `nvidia` | `NVIDIA_API_KEY` (optional: `NVIDIA_BASE_URL`) |
+| GMI Cloud | `gmi` | `GMI_API_KEY` (optional: `GMI_BASE_URL`) |
+| StepFun | `stepfun` | `STEPFUN_API_KEY` (optional: `STEPFUN_BASE_URL`) |
 | Ollama Cloud | `ollama-cloud` | `OLLAMA_API_KEY` |
 | Google Gemini (OAuth) | `google-gemini-cli` | `hermes model` (Google OAuth; optional: `HERMES_GEMINI_PROJECT_ID`) |
 | Google AI Studio | `gemini` | `GOOGLE_API_KEY` (alias: `GEMINI_API_KEY`) |

website/docs/user-guide/features/plugins.md

@@ -93,6 +93,8 @@ Project-local plugins under `./.hermes/plugins/` are disabled by default. Enable
 
 ## What plugins can do
 
+Every `ctx.*` API below is available inside a plugin's `register(ctx)` function.
+
 | Capability | How |
 |-----------|-----|
 | Add tools | `ctx.register_tool(name=..., toolset=..., schema=..., handler=...)` |
@@ -104,6 +106,11 @@ Project-local plugins under `./.hermes/plugins/` are disabled by default. Enable
 | Bundle skills | `ctx.register_skill(name, path)` — namespaced as `plugin:skill`, loaded via `skill_view("plugin:skill")` |
 | Gate on env vars | `requires_env: [API_KEY]` in plugin.yaml — prompted during `hermes plugins install` |
 | Distribute via pip | `[project.entry-points."hermes_agent.plugins"]` |
+| Register a gateway platform (Discord, Telegram, IRC, …) | `ctx.register_platform(name, label, adapter_factory, check_fn, ...)` — see [Adding Platform Adapters](/docs/developer-guide/adding-platform-adapters) |
+| Register an image-generation backend | `ctx.register_image_gen_provider(provider)` — see `plugins/image_gen/openai/` for an example |
+| Register a context-compression engine | `ctx.register_context_engine(engine)` — see [Context Engine Plugins](/docs/developer-guide/context-engine-plugin) |
+| Register a memory backend | Subclass `MemoryProvider` in `plugins/memory/<name>/__init__.py` — see [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) (uses a separate discovery system) |
+| Register an inference backend (LLM provider) | `register_provider(ProviderProfile(...))` in `plugins/model-providers/<name>/__init__.py` — see [Model Provider Plugins](/docs/developer-guide/model-provider-plugin) (uses a separate discovery system) |
 
 ## Plugin discovery
 
@@ -117,6 +124,21 @@ Project-local plugins under `./.hermes/plugins/` are disabled by default. Enable
 
 Later sources override earlier ones on name collision, so a user plugin with the same name as a bundled plugin replaces it.
 
+### Plugin sub-categories
+
+Within each source, Hermes also recognizes sub-category directories that route plugins to specialized discovery systems:
+
+| Sub-directory | What it holds | Discovery system |
+|---|---|---|
+| `plugins/` (root) | General plugins — tools, hooks, slash commands, CLI commands, bundled skills | `PluginManager` (kind: `standalone` or `backend`) |
+| `plugins/platforms/<name>/` | Gateway channel adapters (`ctx.register_platform()`) | `PluginManager` (kind: `platform`, one level deeper) |
+| `plugins/image_gen/<name>/` | Image-generation backends (`ctx.register_image_gen_provider()`) | `PluginManager` (kind: `backend`, one level deeper) |
+| `plugins/memory/<name>/` | Memory providers (subclass `MemoryProvider`) | **Own loader** in `plugins/memory/__init__.py` (kind: `exclusive` — one active at a time) |
+| `plugins/context_engine/<name>/` | Context-compression engines (`ctx.register_context_engine()`) | **Own loader** in `plugins/context_engine/__init__.py` (one active at a time) |
+| `plugins/model-providers/<name>/` | LLM provider profiles (`register_provider(ProviderProfile(...))`) | **Own loader** in `providers/__init__.py` (lazily scanned on first `get_provider_profile()` call) |
+
+User plugins at `~/.hermes/plugins/model-providers/<name>/` and `~/.hermes/plugins/memory/<name>/` override bundled plugins of the same name — last-writer-wins in `register_provider()` / `register_memory_provider()`. Drop a directory in, and it replaces the built-in without any repo edits.
+
 ## Plugins are opt-in
 
 **Every plugin — user-installed, bundled, or pip — is disabled by default.** Discovery finds them (so they show up in `hermes plugins` and `/plugins`), but nothing loads until you add the plugin's name to `plugins.enabled` in `~/.hermes/config.yaml`. This stops anything with hooks or tools from running without your explicit consent.
@@ -163,15 +185,43 @@ Plugins can register callbacks for these lifecycle events. See the **[Event Hook
 
 ## Plugin types
 
-Hermes has three kinds of plugins:
+Hermes has four kinds of plugins:
 
 | Type | What it does | Selection | Location |
 |------|-------------|-----------|----------|
 | **General plugins** | Add tools, hooks, slash commands, CLI commands | Multi-select (enable/disable) | `~/.hermes/plugins/` |
 | **Memory providers** | Replace or augment built-in memory | Single-select (one active) | `plugins/memory/` |
 | **Context engines** | Replace the built-in context compressor | Single-select (one active) | `plugins/context_engine/` |
+| **Model providers** | Declare an inference backend (OpenRouter, Anthropic, …) | Multi-register, picked by `--provider` / `config.yaml` | `plugins/model-providers/` |
 
-Memory providers and context engines are **provider plugins** — only one of each type can be active at a time. General plugins can be enabled in any combination.
+Memory providers and context engines are **provider plugins** — only one of each type can be active at a time. Model providers are also plugins, but many load simultaneously; the user picks one at a time via `--provider` or `config.yaml`. General plugins can be enabled in any combination.
+
+## Pluggable interfaces — where to go for each
+
+The table above shows the four plugin categories, but within "General plugins" the `PluginContext` exposes several distinct extension points — and Hermes also accepts extensions outside the Python plugin system (config-driven backends, shell-hooked commands, external servers, etc.). Use this table to find the right doc for what you want to build:
+
+| Want to add… | How | Authoring guide |
+|---|---|---|
+| A **tool** the LLM can call | Python plugin — `ctx.register_tool()` | [Build a Hermes Plugin](/docs/guides/build-a-hermes-plugin) · [Adding Tools](/docs/developer-guide/adding-tools) |
+| A **lifecycle hook** (pre/post LLM, session start/end, tool filter) | Python plugin — `ctx.register_hook()` | [Hooks reference](/docs/user-guide/features/hooks) · [Build a Hermes Plugin](/docs/guides/build-a-hermes-plugin) |
+| A **slash command** for the CLI / gateway | Python plugin — `ctx.register_command()` | [Build a Hermes Plugin](/docs/guides/build-a-hermes-plugin) · [Extending the CLI](/docs/developer-guide/extending-the-cli) |
+| A **subcommand** for `hermes <thing>` | Python plugin — `ctx.register_cli_command()` | [Extending the CLI](/docs/developer-guide/extending-the-cli) |
+| A bundled **skill** that your plugin ships | Python plugin — `ctx.register_skill()` | [Creating Skills](/docs/developer-guide/creating-skills) |
+| An **inference backend** (LLM provider: OpenAI-compat, Codex, Anthropic-Messages, Bedrock) | Provider plugin — `register_provider(ProviderProfile(...))` in `plugins/model-providers/<name>/` | **[Model Provider Plugins](/docs/developer-guide/model-provider-plugin)** · [Adding Providers](/docs/developer-guide/adding-providers) |
+| A **gateway channel** (Discord / Telegram / IRC / Teams / etc.) | Platform plugin — `ctx.register_platform()` in `plugins/platforms/<name>/` | [Adding Platform Adapters](/docs/developer-guide/adding-platform-adapters) |
+| A **memory backend** (Honcho, Mem0, Supermemory, …) | Memory plugin — subclass `MemoryProvider` in `plugins/memory/<name>/` | [Memory Provider Plugins](/docs/developer-guide/memory-provider-plugin) |
+| A **context-compression strategy** | Context-engine plugin — `ctx.register_context_engine()` | [Context Engine Plugins](/docs/developer-guide/context-engine-plugin) |
+| An **image-generation backend** (DALL·E, SDXL, …) | Backend plugin — `ctx.register_image_gen_provider()` | See bundled examples in `plugins/image_gen/openai/` and `plugins/image_gen/xai/` |
+| A **TTS backend** (any CLI — Piper, VoxCPM, Kokoro, xtts, voice-cloning scripts, …) | Config-driven — declare under `tts.providers.<name>` with `type: command` in `config.yaml` | [TTS setup](/docs/user-guide/features/tts#custom-command-providers) |
+| An **STT backend** (custom whisper binary, local ASR CLI) | Config-driven — set `HERMES_LOCAL_STT_COMMAND` env var to a shell template | [Voice Message Transcription (STT)](/docs/user-guide/features/tts#voice-message-transcription-stt) |
+| **External tools via MCP** (filesystem, GitHub, Linear, Notion, any MCP server) | Config-driven — declare `mcp_servers.<name>` with `command:` / `url:` in `config.yaml`. Hermes auto-discovers the server's tools and registers them alongside built-ins. | [MCP](/docs/user-guide/features/mcp) |
+| **Additional skill sources** (custom GitHub repos, private skill indexes) | CLI — `hermes skills tap add <repo>` | [Skills Hub](/docs/user-guide/features/skills#skills-hub) |
+| **Gateway event hooks** (fire on `gateway:startup`, `session:start`, `agent:end`, `command:*`) | Drop `HOOK.yaml` + `handler.py` into `~/.hermes/hooks/<name>/` | [Event Hooks](/docs/user-guide/features/hooks#gateway-event-hooks) |
+| **Shell hooks** (run a shell command on events — notifications, audit logs, desktop alerts) | Config-driven — declare under `hooks:` in `config.yaml` | [Shell Hooks](/docs/user-guide/features/hooks#shell-hooks) |
+
+:::note
+Not everything is a Python plugin. Some extension surfaces intentionally use **config-driven shell commands** (TTS, STT, shell hooks) so any CLI you already have becomes a plugin without writing Python. Others are **external servers** (MCP) the agent connects to and auto-registers tools from. And some are **drop-in directories** (gateway hooks) with their own manifest format. Pick the right surface for the integration style that fits your use case; the authoring guides in the table above each cover placeholders, discovery, and examples.
+:::
 
 ## NixOS declarative plugins
 

website/sidebars.ts

@@ -209,6 +209,7 @@ const sidebars: SidebarsConfig = {
             'developer-guide/adding-platform-adapters',
             'developer-guide/memory-provider-plugin',
             'developer-guide/context-engine-plugin',
+            'developer-guide/model-provider-plugin',
             'developer-guide/creating-skills',
             'developer-guide/extending-the-cli',
           ],