MCP-Airflow-API

MCP-Airflow-API is an MCP server that connects LLM clients to Apache Airflow clusters using the Model Context Protocol. It bridges natural language interfaces directly to the Airflow REST API, hiding the complexity of raw HTTP endpoints and query parameters. Data engineers, site reliability engineers, and operations teams use this server to inspect pipeline health, manage workflow states, and debug failures through conversational prompts. The server dynamically adapts to both Airflow API v1 (Airflow 2.x) and API v2 (Airflow 3.0+), providing up to 45 tools for managing DAGs, tracking task instances, inspecting execution logs, and modifying connections or variables. It supports pagination and filtering to handle enterprise deployments with large numbers of workflows. Operating over both standard I/O (stdio) and HTTP streaming modes, MCP-Airflow-API allows integration with local desktop assistants like Claude Desktop as well as containerized platforms like OpenWebUI.

Category: Data & Analytics

Tags: airflow, data-pipelines, etl, workflow-orchestration

Visit MCP-Airflow-API

How to install and configure MCP-Airflow-API

1. Ensure Python 3.12 and uv are installed on your system. 2. Open your Claude Desktop configuration file (claude_desktop_config.json). 3. Add the server under the mcpServers object using uvx: json { "mcpServers": { "mcp-airflow-api": { "command": "uvx", "args": ["--python", "3.12", "mcp-airflow-api"], "env": { "AIRFLOW_API_VERSION": "v2", "AIRFLOW_API_BASE_URL": "http://localhost:8080/api", "AIRFLOW_API_USERNAME": "airflow", "AIRFLOW_API_PASSWORD": "airflow" } } } } 4. For remote setups, use streamable-http mode pointing to http://localhost:8000/mcp with an optional Bearer token. 5. Restart Claude Desktop to start managing Airflow pipelines.

What you can do with MCP-Airflow-API

• Trigger, pause, resume, and track the status of specific Apache Airflow DAGs using plain language queries. - Diagnose pipeline failures by retrieving task execution logs, error traces, and runtime XCom values directly. - Monitor overall cluster health, active worker pools, and system configurations across single or multiple Airflow deployments. - Manage Airflow 3.0+ data-aware scheduling assets, connections, and system variables through automated conversational workflows.

Key facts

• Open Source
• https://github.com/call518/MCP-Airflow-API
• Data & Analytics, DevOps, CI/CD & Version Control
• airflow, data-pipelines, etl, workflow-orchestration

Part of MCP Servers

How do I install MCP-Airflow-API?

You can install and run MCP-Airflow-API directly using uvx with Python 3.12 by running uvx --python 3.12 mcp-airflow-api. It can also be built from source using git clone and pip install -e ., or launched as part of a full Docker Compose environment alongside OpenWebUI.

Which Airflow versions are supported by MCP-Airflow-API?

The server supports both Airflow API v1 for Airflow 2.x clusters and Airflow API v2 for Airflow 3.0 and newer. You can switch between them dynamically at runtime by setting the AIRFLOW_API_VERSION environment variable to either v1 or v2 without modifying your MCP tool configurations.

Which MCP clients work with MCP-Airflow-API?

MCP-Airflow-API works with Claude Desktop via stdio transport mode, as well as web interfaces like OpenWebUI. Because it supports both stdio and streamable-http transports with optional Bearer token authentication, it can integrate with any client conforming to the Model Context Protocol.

Is MCP-Airflow-API open source?

Yes, MCP-Airflow-API is an open-source project released under the MIT license. The source code, issue tracker, and contribution guidelines are publicly accessible on GitHub.