us-all/airflow-mcp-server
🔌 MCP Serverus-all
Connect your AI agents to Apache Airflow to manage DAGs, monitor runs, and trigger workflows seamlessly via MCP.
The airflow-mcp-server acts as a specialized interface between AI agents and Apache Airflow, enabling programmatic control over data pipelines. It implements the Model Context Protocol (MCP) to expose seven core tools: listing DAGs, monitoring DAG runs, inspecting task instances, tailing logs, triggering new runs, and clearing existing tasks. A standout feature is the 'dag-health-rollup' aggregation, which provides agents with a high-level summary of pipeline health, reducing the need for multiple granular queries. The server is built to handle Airflow 3.x's /api/v2 endpoints using JWT authentication via SimpleAuthManager. Security is prioritized through a read-only default configuration, with write-gated actions like triggering or clearing tasks requiring explicit permission. This architecture allows developers to integrate complex workflow orchestration into agentic workflows without exposing the entire Airflow backend, ensuring a secure and controlled interaction layer for automated data engineering.
💡Highlights
- ├─7 tools for DAG management
- ├─Supports Airflow 3.x & 2.x
- └─Read-only default security
🎯For
- ├─Data Engineers
- ├─AI Agents Developers
- └─MLOps Engineers