Langflow MCP Server

MCP Server for interacting with the Langflow API via a natural language interface.

Overview

This server exposes Langflow functionalities as "tools" that can be understood and invoked by a Large Language Model (LLM) through an MCP Client. It allows for programmatic interaction with flows, projects, components, and more.

Related MCP server: mcp-cli-catalog

Project Structure

• src/langflowmcp_server/app.py: The main FastMCP application entry point. Manages server lifecycle and session handling.

• src/langflowmcp_server/langflow_api_client.py: A dedicated client for all communication with the Langflow REST API.

• src/langflowmcp_server/langflow_models.py: Pydantic models for all Langflow API data structures.

• src/langflowmcp_server/tools/: Directory containing the tool implementations, organized by functionality (e.g., flows.py, execution.py).

Setup

1. Clone the repository.

2. Ensure Python 3.10+ and uv are installed.

3. Create a virtual environment:

   uv venv
   source .venv/bin/activate

4. Install dependencies:

   uv pip install -e .

5. Set up your Langflow connection details in a .env file (see .env.example). You will need your Langflow URL and an API key.

Running the Server

You can run the server directly for stdio-based communication:

python -m src.langflowmcp_server.app