Connect LangChain (Python) to Slack, Gmail, Notion and GitHub through a remote Weavz Model Context Protocol (MCP) server. Use the native MCP adapter, then try the app setup recipes with your configured connections.
App setup recipes:
- Slack MCP with LangChain Python
- Gmail MCP with LangChain Python
- Notion MCP with LangChain Python
- GitHub MCP with LangChain Python
The framework consumes app tools from your configured Weavz MCP server. App recipes require your own app connections; automated checks use synthetic records.
MCP app and framework directory · All examples · Weavz MCP guide
Uses langchain-mcp-adapters to expose the selected Weavz server's tools as LangChain tools. Tool Mode/Code Mode are configured server-side; the local --mode flag validates the expected names and does not switch a live server.
Requires Python 3.11 or 3.12, uv, a custom Weavz Streamable HTTP URL and MCP access token. Do not substitute a Weavz REST API key. The hosted personal /mcp/weavz connector uses OAuth.
uv sync --locked
uv run python main.py --fixture --mode tools
uv run python main.py --fixture --mode code
uv run pytestThe fixture is a separate loopback MCP SDK server with bearer auth and synthetic data. No model or live Weavz call is made. For live discovery, copy .env.example to .env and run uv run --env-file .env python main.py --mode tools or --mode code. For an optional LangChain agent run, connect an app with relevant data, set WEAVZ_MODEL and OPENAI_API_KEY, then add --prompt "Find the latest record matching the task you connected". Each tool call uses the adapter-managed HTTP lifecycle.