feat: Add MCP-backed dynamic tool fetching to Python SDK #39
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR implements MCP (Model Context Protocol) integration for dynamic tool discovery in the Python SDK, achieving feature parity with the Node SDK. The changes enable runtime tool fetching from the StackOne API without requiring SDK updates.
Key changes:
- Added RPC-backed tool execution through the
/actions/rpcendpoint via new_StackOneRpcToolclass - Implemented MCP client integration with pagination support using
streamablehttp_clienttransport - Added comprehensive test suite with 12 tests covering account, provider, and action filtering scenarios
Reviewed Changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
tests/test_toolset_mcp.py |
New comprehensive test suite with 12 tests organized into TestAccountFiltering and TestProviderAndActionFiltering classes |
tests/test_toolset.py |
Removed obsolete non-MCP fetch_tools tests, cleaned up imports |
stackone_ai/toolset.py |
Core MCP implementation including _StackOneRpcTool, _fetch_mcp_tools, thread-safe async wrapper, and enhanced fetch_tools() method |
stackone_ai/feedback/tool.py |
Added type ignore comments for Pydantic field validators and improved code formatting |
README.md |
Updated documentation with MCP feature information and installation instructions |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| try: | ||
| from mcp import types as mcp_types | ||
| from mcp.client.session import ClientSession | ||
| from mcp.client.streamable_http import streamablehttp_client |
There was a problem hiding this comment.
The import path mcp.client.streamable_http appears to be incorrect. Based on the MCP Python SDK package structure, this should be mcp.client.streamablehttp (without the underscore). The correct import should be from mcp.client.streamablehttp import streamablehttp_client.
Suggested change
| from mcp.client.streamable_http import streamablehttp_client | |
| from mcp.client.streamablehttp import streamablehttp_client |
| def runner() -> None: | ||
| try: | ||
| result["value"] = asyncio.run(awaitable) | ||
| except BaseException as exc: # pragma: no cover - surfaced in caller context |
There was a problem hiding this comment.
Except block directly handles BaseException.
Suggested change
| except BaseException as exc: # pragma: no cover - surfaced in caller context | |
| except Exception as exc: # pragma: no cover - surfaced in caller context |
ryoppippi
force-pushed
the
ENG-11393-add-fetch-tools-in-python-sdk
branch
from
f195fb5 to
764c149
Compare
Add MCP (Model Context Protocol) integration to enable dynamic tool fetching at runtime, matching the Node SDK functionality. This allows users to pull the latest tool definitions from the StackOne API without SDK updates. Key Features: - Dynamic tool discovery via MCP server endpoint - _StackOneRpcTool class for RPC-backed execution at /actions/rpc - Account filtering with set_accounts() and account_ids parameter - Provider filtering (case-insensitive, prefix-based) - Action filtering with glob pattern support - Thread-safe async execution wrapper (_run_async) - Proper authentication with Basic Auth headers Implementation Details: - MCP client integration with streamablehttp transport - Tool catalog fetching with pagination support (cursor-based) - Schema normalization for proper nullable field handling - RPC payload construction with body/headers/path/query support - Account scoping via x-account-id header - User-Agent tracking for SDK version telemetry Dependencies: - Requires optional 'mcp' extra: pip install 'stackone-ai[mcp]' - Raises ToolsetConfigError if MCP dependencies not available The implementation maintains API parity with the Node SDK's stackone.mcp-fetch functionality while following Python idioms.
Add 12 new tests covering MCP-backed tool fetching to match Node SDK test coverage in stackone.mcp-fetch.spec.ts. Tests are organized into logical groups for account and provider/action filtering scenarios. Test Coverage: Account Filtering (TestAccountFiltering): - fetch_tools_with_single_account_id: Single account filtering - fetch_tools_with_multiple_account_ids: Multiple account filtering - fetch_tools_with_set_accounts: Using set_accounts() method - fetch_tools_account_ids_override_set_accounts: Parameter precedence - fetch_tools_with_constructor_account_id: Constructor-based account - fetch_tools_with_empty_account_ids: Empty list handling Provider/Action Filtering (TestProviderAndActionFiltering): - fetch_tools_with_provider_filter: Single provider filtering - fetch_tools_with_multiple_providers: Multiple provider filtering - fetch_tools_with_action_glob_pattern: Glob pattern matching - fetch_tools_with_exact_action_match: Exact action name matching - fetch_tools_with_provider_and_action_filters: Combined filtering - fetch_tools_with_exclusion_pattern: Negative glob patterns Removed: - Deleted old fetch_tools tests that used non-MCP approach - These tests were testing the old implementation that loaded from OpenAPI specs instead of the MCP server The test suite uses comprehensive mocking of MCP client components: - ClientSession for MCP protocol communication - streamablehttp_client for HTTP transport - Tool list responses with pagination support - Proper async context manager handling All tests verify: - Correct tool count after filtering - Proper tool presence/absence - Account ID preservation in tool context - Filter precedence and interaction
ryoppippi
force-pushed
the
ENG-11393-add-fetch-tools-in-python-sdk
branch
from
764c149 to
1cecc8a
Compare
ryoppippi
force-pushed
the
ENG-11393-add-fetch-tools-in-python-sdk
branch
from
1cecc8a to
a127c6c
Compare
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
NicolasBelissent
approved these changes
Nov 12, 2025
| import threading | ||
| import warnings | ||
| from typing import Any | ||
| from collections.abc import Coroutine |
Contributor
There was a problem hiding this comment.
never heard of these they good?
Member
Author
There was a problem hiding this comment.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implements
fetch_tools()method with MCP-backed dynamic tool discovery, matching the Node SDK implementation.Related: StackOneHQ/stackone-ai-node#114