mcp-openapi-proxy 0.3.0__tar.gz → 0.3.2__tar.gz

{mcp_openapi_proxy-0.3.0 → mcp_openapi_proxy-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mcp-openapi-proxy
-Version: 0.3.0
+Version: 0.3.2
 Summary: MCP server for exposing OpenAPI specifications as MCP tools.
 Author-email: Matthew Hand <11550632+matthewhand@users.noreply.github.com>
 Requires-Python: >=3.10
@@ -175,27 +175,32 @@ The example configurations below were exercised against the live APIs, and the p
 
 | Agent CLI | Model used (live test) | MCP attach mechanism | Tool calls | Prompts/Resources surfaced to model? |
 |---|---|---|---|---|
-| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | unknown ‡ |
-| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: ❌ † |
-| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ † |
-| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json`, clean workspace | ✅ native | prompts: ❌ · resources: ✅ (`access_mcp_resource`) † |
-| opencode | `orchestration` group via local LiteLLM gateway | `~/.config/opencode/opencode.json` `mcp` | ✅ native | unknown ‡ |
-| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads (writes flaky) | prompts: ❌ · resources: ❌ (tools-only) † |
+| **opencode** | (CLI default) | `~/.config/opencode/opencode.json` `mcp` | ✅ native | **prompts: ✅ (slash) · resources: ✅** — most complete ‖ |
+| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | prompts: ❌ (no prompt meta-tools) · resources: ✅ (`read_mcp_resource`) ‖ |
+| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json` | ✅ native | prompts: ❌ (no prompt mechanism) · resources: ✅ (`access_mcp_resource`) ‖ |
+| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native (live invoke auth-blocked) | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ (no client support) ‖ |
+| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: interactive `@` only (neither reaches the model headless) ‖ |
+| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads | prompts: ❌ · resources: ❌ (tools-only client) ‖ |
 | agy | — | — | ❌ headless cannot enable MCP | n/a |
-| letta cloud | Letta Cloud default | streamable-HTTP MCP URL (`/mcp add --transport http` + bearer) | ✅ remote (stdio rejected) | unknown ‡ |
-| letta (self-hosted ≤0.11.x) | `agent` group via local LiteLLM gateway | stdio via `PUT /v1/tools/mcp/servers` | ✅ native | unknown ‡ |
-
-> **Prompts/resources column — read this.** The original 0.2.0 sweep ran while the
-> server advertised prompts/resources **only when `ENABLE_PROMPTS`/`ENABLE_RESOURCES`
-> were set — they defaulted OFF.** Per the MCP spec a client won't call
-> `prompts/list`/`resources/list` unless the capability is advertised, so those early
-> results measured the *server's* default, not the clients. **All prior prompt/resource
-> findings are therefore voided.** This release defaults advertising **on**.
+| letta (cloud / self-hosted) | Letta Cloud / `PUT /v1/tools/mcp/servers` | streamable-HTTP / stdio | ✅ (per prior sweep) | unknown — needs a running Letta server to test ‡ |
+
+> **How to read the prompts/resources column.** MCP has three surfaces — tools, prompts,
+> resources. **mcp-openapi-proxy serves all three, advertised by default since 0.3.0.**
+> Whether they reach the model is up to the *client*, and that varies:
 >
-> - **† re-verified 2026-06-14** with advertising on, against the real client binary. Genuine, uneven: **tools** everywhere; **prompts→model** Qwen (slash) & Gemini (interactive only); **resources→model** Kilocode. Vibe is tools-only (confirmed in source).
-> - **‡ unknown** — not yet re-tested under advertising-on (opencode not installed here; Letta needs a running server + gateway; Codex parked). Prior ❌ values are not carried forward.
+> - **‖ re-verified 2026-06-14 against the published 0.3.0 release** with **no** flags set
+>   (validating the default-on advertising), driving each **real client binary**.
+> - **Tools** work on every client tested. **Prompts→model**: opencode & Qwen (slash
+>   commands); Gemini interactive-only. **Resources→model**: opencode, Codex, Kilocode.
+>   **opencode is the only client that surfaces all three.** Vibe is tools-only.
+> - **‡ unknown** — Letta can't be exercised without standing up a Letta server + model;
+>   left untested rather than guessed.
+> - Every cell on the proxy side was confirmed via a raw stdio handshake (`initialize`
+>   advertises tools+prompts+resources; `prompts/get` and `resources/read` return content).
+>   The remaining ❌ are **client-side** gaps, not proxy limitations.
 >
-> Tool-calling (the `Tool calls` column) was unaffected by the advertising default and stands as originally verified.
+> *(All pre-0.3.0 prompt/resource findings were voided — they were measured while
+> advertising defaulted off, so they reflected the server default, not the clients.)*
 
 Minimal sanitized configs per client (the no-auth Glama spec is used as the smallest working example; substitute your own spec URL and `$YOUR_KEY` as needed):
 

{mcp_openapi_proxy-0.3.0 → mcp_openapi_proxy-0.3.2}/README.md

@@ -153,27 +153,32 @@ The example configurations below were exercised against the live APIs, and the p
 
 | Agent CLI | Model used (live test) | MCP attach mechanism | Tool calls | Prompts/Resources surfaced to model? |
 |---|---|---|---|---|
-| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | unknown ‡ |
-| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: ❌ † |
-| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ † |
-| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json`, clean workspace | ✅ native | prompts: ❌ · resources: ✅ (`access_mcp_resource`) † |
-| opencode | `orchestration` group via local LiteLLM gateway | `~/.config/opencode/opencode.json` `mcp` | ✅ native | unknown ‡ |
-| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads (writes flaky) | prompts: ❌ · resources: ❌ (tools-only) † |
+| **opencode** | (CLI default) | `~/.config/opencode/opencode.json` `mcp` | ✅ native | **prompts: ✅ (slash) · resources: ✅** — most complete ‖ |
+| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | prompts: ❌ (no prompt meta-tools) · resources: ✅ (`read_mcp_resource`) ‖ |
+| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json` | ✅ native | prompts: ❌ (no prompt mechanism) · resources: ✅ (`access_mcp_resource`) ‖ |
+| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native (live invoke auth-blocked) | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ (no client support) ‖ |
+| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: interactive `@` only (neither reaches the model headless) ‖ |
+| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads | prompts: ❌ · resources: ❌ (tools-only client) ‖ |
 | agy | — | — | ❌ headless cannot enable MCP | n/a |
-| letta cloud | Letta Cloud default | streamable-HTTP MCP URL (`/mcp add --transport http` + bearer) | ✅ remote (stdio rejected) | unknown ‡ |
-| letta (self-hosted ≤0.11.x) | `agent` group via local LiteLLM gateway | stdio via `PUT /v1/tools/mcp/servers` | ✅ native | unknown ‡ |
-
-> **Prompts/resources column — read this.** The original 0.2.0 sweep ran while the
-> server advertised prompts/resources **only when `ENABLE_PROMPTS`/`ENABLE_RESOURCES`
-> were set — they defaulted OFF.** Per the MCP spec a client won't call
-> `prompts/list`/`resources/list` unless the capability is advertised, so those early
-> results measured the *server's* default, not the clients. **All prior prompt/resource
-> findings are therefore voided.** This release defaults advertising **on**.
+| letta (cloud / self-hosted) | Letta Cloud / `PUT /v1/tools/mcp/servers` | streamable-HTTP / stdio | ✅ (per prior sweep) | unknown — needs a running Letta server to test ‡ |
+
+> **How to read the prompts/resources column.** MCP has three surfaces — tools, prompts,
+> resources. **mcp-openapi-proxy serves all three, advertised by default since 0.3.0.**
+> Whether they reach the model is up to the *client*, and that varies:
 >
-> - **† re-verified 2026-06-14** with advertising on, against the real client binary. Genuine, uneven: **tools** everywhere; **prompts→model** Qwen (slash) & Gemini (interactive only); **resources→model** Kilocode. Vibe is tools-only (confirmed in source).
-> - **‡ unknown** — not yet re-tested under advertising-on (opencode not installed here; Letta needs a running server + gateway; Codex parked). Prior ❌ values are not carried forward.
+> - **‖ re-verified 2026-06-14 against the published 0.3.0 release** with **no** flags set
+>   (validating the default-on advertising), driving each **real client binary**.
+> - **Tools** work on every client tested. **Prompts→model**: opencode & Qwen (slash
+>   commands); Gemini interactive-only. **Resources→model**: opencode, Codex, Kilocode.
+>   **opencode is the only client that surfaces all three.** Vibe is tools-only.
+> - **‡ unknown** — Letta can't be exercised without standing up a Letta server + model;
+>   left untested rather than guessed.
+> - Every cell on the proxy side was confirmed via a raw stdio handshake (`initialize`
+>   advertises tools+prompts+resources; `prompts/get` and `resources/read` return content).
+>   The remaining ❌ are **client-side** gaps, not proxy limitations.
 >
-> Tool-calling (the `Tool calls` column) was unaffected by the advertising default and stands as originally verified.
+> *(All pre-0.3.0 prompt/resource findings were voided — they were measured while
+> advertising defaulted off, so they reflected the server default, not the clients.)*
 
 Minimal sanitized configs per client (the no-auth Glama spec is used as the smallest working example; substitute your own spec URL and `$YOUR_KEY` as needed):
 

{mcp_openapi_proxy-0.3.0 → mcp_openapi_proxy-0.3.2}/mcp_openapi_proxy/server_fastmcp.py

@@ -36,6 +36,55 @@ spec = None  # Global spec for resources
 # callable via call_function.
 _FUNCTION_OPERATIONS: Dict[str, Dict] = {}
 
+
+# --- Native MCP prompts & resources for FastMCP/simple mode ---
+# Previously simple mode advertised the prompts/resources capabilities (FastMCP
+# does so automatically) but registered none, so prompts/list & resources/list
+# came back EMPTY. Register them natively here for parity with the low-level
+# server (same names/content): prompts summarize_spec + whimsical_blog, and the
+# spec_file resource.
+
+def _spec_as_json() -> str:
+    """Serialize the OpenAPI spec for the spec_file resource.
+
+    Serves the in-memory spec that run_simple_server preloads once; only
+    fetches as a lazy fallback. This avoids a live network re-fetch on every
+    resources/read, respecting the live-first-once spec lifecycle (#28).
+    Uses default=str so YAML datetime example values don't crash serialization.
+    """
+    global spec
+    if spec is None:
+        spec_url = os.environ.get("OPENAPI_SPEC_URL")
+        if not spec_url:
+            return json.dumps({"error": "OPENAPI_SPEC_URL is not configured"})
+        spec = fetch_openapi_spec(spec_url)
+    if not spec:
+        return json.dumps({"error": "Failed to fetch OpenAPI spec"})
+    if isinstance(spec, str):
+        return spec
+    return json.dumps(spec, indent=2, default=str)
+
+
+@mcp.resource("file:///openapi_spec.json", name="spec_file",
+              description="The raw OpenAPI specification JSON", mime_type="application/json")
+def _spec_file_resource() -> str:
+    """Serve the configured OpenAPI spec as a native MCP resource."""
+    return _spec_as_json()
+
+
+@mcp.prompt(name="summarize_spec", description="Summarizes the OpenAPI specification")
+def _summarize_spec_prompt() -> str:
+    return ("This OpenAPI spec defines endpoints, parameters, and responses—a "
+            "blueprint for developers to integrate effectively.")
+
+
+@mcp.prompt(name="whimsical_blog", description="A whimsical WordPress blog-post starter inspired by this API")
+def _whimsical_blog_prompt() -> str:
+    return ("Once upon a JSON, in a land of tilde keys and sticky semicolons, a pet AI "
+            "chatbot discovered it could whisper to WordPress through a magic OpenAPI proxy. "
+            "✨ Write the next whimsical chapter.")
+
+
 @mcp.tool()
 def list_functions(*, env_key: str = "OPENAPI_SPEC_URL") -> str:
     """Lists available functions derived from the OpenAPI specification."""
@@ -243,7 +292,8 @@ def call_function(*, function_name: str, parameters: Optional[Dict] = None, env_
             spec_local = json.loads(spec_local)
         if spec_local is None:
             return json.dumps({"error": "Failed to fetch OpenAPI spec"})
-        return json.dumps(spec_local, indent=2)
+        # default=str: YAML datetime example values aren't JSON-serializable
+        return json.dumps(spec_local, indent=2, default=str)
     if function_name == "list_prompts":
         return json.dumps([{"name": "summarize_spec", "description": "Summarizes the purpose of the OpenAPI specification", "arguments": []}])
     if function_name == "get_prompt":

{mcp_openapi_proxy-0.3.0 → mcp_openapi_proxy-0.3.2}/mcp_openapi_proxy.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mcp-openapi-proxy
-Version: 0.3.0
+Version: 0.3.2
 Summary: MCP server for exposing OpenAPI specifications as MCP tools.
 Author-email: Matthew Hand <11550632+matthewhand@users.noreply.github.com>
 Requires-Python: >=3.10
@@ -175,27 +175,32 @@ The example configurations below were exercised against the live APIs, and the p
 
 | Agent CLI | Model used (live test) | MCP attach mechanism | Tool calls | Prompts/Resources surfaced to model? |
 |---|---|---|---|---|
-| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | unknown ‡ |
-| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: ❌ † |
-| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ † |
-| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json`, clean workspace | ✅ native | prompts: ❌ · resources: ✅ (`access_mcp_resource`) † |
-| opencode | `orchestration` group via local LiteLLM gateway | `~/.config/opencode/opencode.json` `mcp` | ✅ native | unknown ‡ |
-| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads (writes flaky) | prompts: ❌ · resources: ❌ (tools-only) † |
+| **opencode** | (CLI default) | `~/.config/opencode/opencode.json` `mcp` | ✅ native | **prompts: ✅ (slash) · resources: ✅** — most complete ‖ |
+| Codex | `gpt-5-codex` (OpenAI API) | `codex exec -c mcp_servers.*` | ✅ native | prompts: ❌ (no prompt meta-tools) · resources: ✅ (`read_mcp_resource`) ‖ |
+| Kilocode | `kilo-auto/free` | global `settings/mcp_settings.json` | ✅ native | prompts: ❌ (no prompt mechanism) · resources: ✅ (`access_mcp_resource`) ‖ |
+| Qwen | `agent` group via local LiteLLM gateway | project `.qwen/settings.json` | ✅ native (live invoke auth-blocked) | prompts: ✅ (slash `/summarize_spec`) · resources: ❌ (no client support) ‖ |
+| Gemini | Google OAuth free tier (CLI default model) | project `.gemini/settings.json` `mcpServers` | ✅ native | prompts: interactive slash only · resources: interactive `@` only (neither reaches the model headless) ‖ |
+| Vibe | `mistral-medium-3.5` | `~/.vibe/config.toml` `[[mcp_servers]]` | ✅ discovery + reads | prompts: ❌ · resources: ❌ (tools-only client) ‖ |
 | agy | — | — | ❌ headless cannot enable MCP | n/a |
-| letta cloud | Letta Cloud default | streamable-HTTP MCP URL (`/mcp add --transport http` + bearer) | ✅ remote (stdio rejected) | unknown ‡ |
-| letta (self-hosted ≤0.11.x) | `agent` group via local LiteLLM gateway | stdio via `PUT /v1/tools/mcp/servers` | ✅ native | unknown ‡ |
-
-> **Prompts/resources column — read this.** The original 0.2.0 sweep ran while the
-> server advertised prompts/resources **only when `ENABLE_PROMPTS`/`ENABLE_RESOURCES`
-> were set — they defaulted OFF.** Per the MCP spec a client won't call
-> `prompts/list`/`resources/list` unless the capability is advertised, so those early
-> results measured the *server's* default, not the clients. **All prior prompt/resource
-> findings are therefore voided.** This release defaults advertising **on**.
+| letta (cloud / self-hosted) | Letta Cloud / `PUT /v1/tools/mcp/servers` | streamable-HTTP / stdio | ✅ (per prior sweep) | unknown — needs a running Letta server to test ‡ |
+
+> **How to read the prompts/resources column.** MCP has three surfaces — tools, prompts,
+> resources. **mcp-openapi-proxy serves all three, advertised by default since 0.3.0.**
+> Whether they reach the model is up to the *client*, and that varies:
 >
-> - **† re-verified 2026-06-14** with advertising on, against the real client binary. Genuine, uneven: **tools** everywhere; **prompts→model** Qwen (slash) & Gemini (interactive only); **resources→model** Kilocode. Vibe is tools-only (confirmed in source).
-> - **‡ unknown** — not yet re-tested under advertising-on (opencode not installed here; Letta needs a running server + gateway; Codex parked). Prior ❌ values are not carried forward.
+> - **‖ re-verified 2026-06-14 against the published 0.3.0 release** with **no** flags set
+>   (validating the default-on advertising), driving each **real client binary**.
+> - **Tools** work on every client tested. **Prompts→model**: opencode & Qwen (slash
+>   commands); Gemini interactive-only. **Resources→model**: opencode, Codex, Kilocode.
+>   **opencode is the only client that surfaces all three.** Vibe is tools-only.
+> - **‡ unknown** — Letta can't be exercised without standing up a Letta server + model;
+>   left untested rather than guessed.
+> - Every cell on the proxy side was confirmed via a raw stdio handshake (`initialize`
+>   advertises tools+prompts+resources; `prompts/get` and `resources/read` return content).
+>   The remaining ❌ are **client-side** gaps, not proxy limitations.
 >
-> Tool-calling (the `Tool calls` column) was unaffected by the advertising default and stands as originally verified.
+> *(All pre-0.3.0 prompt/resource findings were voided — they were measured while
+> advertising defaulted off, so they reflected the server default, not the clients.)*
 
 Minimal sanitized configs per client (the no-auth Glama spec is used as the smallest working example; substitute your own spec URL and `$YOUR_KEY` as needed):
 

{mcp_openapi_proxy-0.3.0 → mcp_openapi_proxy-0.3.2}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "mcp-openapi-proxy"
 requires-python = ">=3.10"
-version = "0.3.0"
+version = "0.3.2"
 description = "MCP server for exposing OpenAPI specifications as MCP tools."
 readme = "README.md"
 authors = [