mcp-server-apache-airflow
Author: yangkyeongmo
Description: Model Context Protocol (MCP) server that wraps the Apache Airflow REST API so MCP-compatible clients can manage DAGs, runs, tasks, variables, connections, pools, etc. through a unified interface.
Stars: 72
Forks: 20
License: MIT License
Category: Open Source
Overview
Installation
1. Prerequisites:
• Python 3.8+
• An existing Apache Airflow installation (>=2.5)
2. Install the MCP Airflow server package from PyPI:
pip install mcp-server-apache-airflow3. Enable the provider inside Airflow:
In your `airflow.cfg` or via environment variable, add the plug-in to the plugins folder if required, e.g.
export AIRFLOW__CORE__PLUGINS_FOLDER=$AIRFLOW_HOME/plugins The package automatically exposes its plugins after installation; restart the Airflow web-server and scheduler.
4. Configuration:
• Set the MCP endpoint and token in an Airflow connection or via environment variables, e.g.
export MCP_API_URL="https://<your-mcp-host>/api"
export MCP_API_TOKEN="<personal-access-token>" • Optionally configure connection id `mcp_default` in the Airflow UI (type: HTTP) that stores the same URL and token.
5. Verify:
• In the Airflow UI, navigate to Admin ➜ Connections and ensure `mcp_default` exists.
• Run the example DAG that ships with the package (`example_mcp_dag`) to validate connectivity.
License: MIT License
Updated 7/30/2025