Airflow MCP

⚠️ REPOSITORY MOVED - NO LONGER MAINTAINED HERE

This repository has been transferred to new ownership and is no longer actively maintained in this location.

This repository and all associated open-source packages have been moved to a new GitHub organization.

New Location:https://github.com/ponderedw

Please visithttps://github.com/ponderedwto:

You'll need to configure Claude Desktop to connect to your Airflow instance. If you haven't configured Claude Desktop for use with MCP before, we recommend following theClaude Desktop documentation.

{ "mcpServers": { "airflow_mcp": { "command": "docker", "args": ["run", "-i", "--rm", "-e", "airflow_api_url", "-e", "airflow", "-e", "airflow", "hipposysai/airflow-mcp:latest"], "env": { "airflow_api_url": "http://host.docker.internal:8088/api/v1", "airflow_username": "airflow", "airflow_password": "airflow" } } } }

git clone https://github.com/hipposys-ltd/airflow-mcp

If you don't have a running Airflow environment, start one with:

This will start an Airflow instance on port 8088, with usernameairflowand passwordairflow.

You can access Airflow athttp://localhost:8088/and see multiple DAGs configured:

You'll need to configure Claude Desktop to connect to your Airflow instance. If you haven't configured Claude Desktop for use with MCP before, we recommend following theClaude Desktop documentation.

{ "mcpServers": { "airflow_mcp": { "command": "docker", "args": ["run", "-i", "--rm", "-e", "airflow_api_url", "-e", "airflow", "-e", "airflow", "hipposysai/airflow-mcp:latest"], "env": { "airflow_api_url": "http://host.docker.internal:8088/api/v1", "airflow_username": "airflow", "airflow_password": "airflow" } } } }

Test your setup by asking Claude: "What DAGs do we have in our Airflow cluster?"

- For AWS Bedrock:LLM_MODEL_ID=bedrock:...
- For Anthropic:LLM_MODEL_ID=anthropic:...
- For OpenAI:LLM_MODEL_ID=openai:...

- AWS credentials for Bedrock
- ANTHROPIC_API_KEYfor Claude
- OPENAI_API_KEYfor ChatGPT

airflow_api_url=your_airflow_api_url airflow_username=your_airflow_username airflow_password=your_airflow_password

- With bundled Airflow:just project
- With existing Airflow:just project_no_airflow

Try it out by asking "How many DAGs failed today?" in the Chat UI

- "What DAGs do we have in our Airflow cluster?"
- "Identify all DAGs with failed status in their most recent execution and trigger a new run for each one"
- "What operators are used by the transform_forecast_attendance DAG?"
- "Has the transform_forecast_attendance DAG ever completed successfully?"

You can usefrom langchain_mcp_adapters.client import MultiServerMCPClientin order to add our Airflow MCP as one of your tools in your LangChain app.

Option 1: Separate Container (SSE Transport)

If you run our Airflow MCP as a separate container, use SSE transport:

mcp_host = os.environ.get('mcp_host', 'mcp_sse_server:8000') mcps = { "AirflowMCP": { "url": f"http://{mcp_host}/sse", "transport": "sse", "headers": {"Authorization": f"""Bearer { os.environ.get('MCP_TOKEN')}"""} } }

Option 2: Embedded Server (STDIO Transport)

If you want to run our MCP server as part of the LangChain code, without any outside code, use STDIO and make sure you install our library first (airflow-mcp-hipposys = "0.1.0a11"):

mcps = { "AirflowMCP": { 'command': "python", 'args': ["-m", "airflow_mcp_hipposys.mcp_airflow"], "transport": "stdio", 'env': {k: v for k, v in { 'AIRFLOW_ASSISTENT_AI_CONN': os.getenv( 'AIRFLOW_ASSISTENT_AI_CONN'), 'airflow_api_url': os.getenv('airflow_api_url'), 'airflow_username': os.getenv('airflow_username'), 'airflow_password': os.getenv('airflow_password'), 'AIRFLOW_INSIGHTS_MODE': os.getenv('AIRFLOW_INSIGHTS_MODE'), 'POST_MODE': os.getenv('POST_MODE'), 'TRANSPORT_TYPE': 'stdio', '_AIRFLOW_WWW_USER_USERNAME': os.getenv('_AIRFLOW_WWW_USER_USERNAME'), '_AIRFLOW_WWW_USER_PASSWORD': os.getenv('_AIRFLOW_WWW_USER_PASSWORD') }.items() if v is not None} } }

Then in both cases, pass it to the tools:

client = MultiServerMCPClient(mcps) tools = await client.get_tools()

We enthusiastically invite the community to contribute to this open-source initiative! Whether you're interested in:

- Adding new features
- Improving documentation
- Enhancing compatibility with different LLM providers
- Reporting bugs
- Suggesting improvements

Please feel free to submit pull requests or open issues on our GitHub repository.

Scout's official MCP pipes error, trace and metric data from production to your AI agent

Interact with your MLOps and LLMOps pipelines through your ZenML MCP server

Paid remote MCP for CLI tool MCP, structured receipts, usage logs, and audit-ready evidence for agent and CI workflows.

Search and run capsules, execute pipelines, and manage data assets on the Code Ocean platform.

Remote MCP server that gives your AI agent a dead-man's-switch: it calls create_heartbeat, pings on schedule, and emails you the instant an agent or cron loop stops checking in.

MCP server for the Enoch control plane: dispatch gates, evidence sync, provenance, and quality gates for long-running autonomous AI work.

MCP server for Hatchet - let AI agents observe and operate your workflows: runs, logs, workers, metrics, plus trigger/cancel/replay

simple mcp server for Airflow 3 (API version 2)

AI-powered Jenkins pipeline intelligence platform with natural language interface. Provides comprehensive pipeline analysis, failure prediction, optimization suggestions, and automated Jenkinsfile reconstruction using Model Context Protocol (MCP) integration.