mcp-airflow-simple

simple mcp server for Airflow 3 (API version 2)

AModel Context Protocol (MCP) serverfor Apache Airflow 3 that provides essential tools for DAG management, monitoring, debugging, and connection testing through the Airflow REST API v2.

it will return a token, copy the token and paste it to the .env file

make sure your airflow is running and accessible at the configured URL

curl -X POST "{your_ariflow_url}/auth/token" -H "Content-Type: application/json" -d '{"username":"{your_airflow_username}","password":"{your_airflow_password}"}'

curl -X POST "http://localhost:8080/auth/token" -H "Content-Type: application/json" -d '{"username":"airflow","password":"airflow"}'

{ "mcpServers": { "airflow": { "command": "python", "args": ["c:\\{path_to_your_folder}\\mcp-airflow-simple\\server.py"], "env": { "GIT_AUTO_UPDATE": "true" } } } }

- List all DAGs with filtering options
- Get tasks within a specific DAG
- Trigger DAG runs with optional configuration
- Clear/retry failed DAG runs

- Check DAG run history and status
- View task instances for specific runs
- Get aggregate DAG statistics

- Retrieve task execution logs
- Check DAG import/parsing errors

- List all Airflow connections
- Get connection details
- Test connection accessibility

- Monitor Airflow Scheduler, Metadatabase, Triggerer, and DagProcessor status
-

Clone or navigate to the project directory:

Configure environment variables:Edit the.envfile with your Airflow instance details:

airflow_baseurl=http://localhost:8080 airflow_api_url=http://localhost:8080/api/v2 airflow_username=airflow airflow_password=airflow airflow_jwt_token=your_jwt_token_here

The server supports two authentication methods:
- JWT Token (Preferred): Setairflow_jwt_tokenin.env
- Basic Auth (Fallback): Usesairflow_usernameandairflow_password

The server will automatically use JWT if available, otherwise fall back to basic authentication.

{ "only_active": false, "limit": 100 }

{ "dag_id": "example_dag" }

{ "dag_id": "example_dag", "conf": {"key": "value"}, "logical_date": "2026-01-05T00:00:00Z" }

Clear/retry a DAG run (resets failed tasks).

{ "dag_id": "example_dag", "dag_run_id": "manual__2026-01-05T00:00:00+00:00", "dry_run": false }

{ "dag_id": "example_dag", "is_paused": true }

Get DAG run history with optional state filtering.

{ "dag_id": "example_dag", "state": "failed", "limit": 25 }

Get task instances for a specific DAG run.

{ "dag_id": "example_dag", "dag_run_id": "manual__2026-01-05T00:00:00+00:00" }

Get execution logs for a specific task instance.

{ "dag_id": "example_dag", "dag_run_id": "manual__2026-01-05T00:00:00+00:00", "task_id": "example_task", "try_number": 1 }

{ "connection_id": "postgres_default" }

{ "connection_id": "postgres_default" }

Check Airflow system health (includes Metadatabase, Scheduler, Triggerer, and DagProcessor).

The server runs as a stdio-based MCP server:

To use this server with MCP clients like Claude Desktop, add to your MCP configuration:

Windows(%APPDATA%\Claude\claude_desktop_config.json):

{ "mcpServers": { "airflow": { "command": "python", "args": ["c:\\{path_to_your_folder}\\mcp-airflow\\server.py"], "env": { "airflow_api_url": "http://localhost:8080/api/v2", "airflow_jwt_token": "your_token_here" } } } }

macOS/Linux(~/Library/Application Support/Claude/claude_desktop_config.json):

{ "mcpServers": { "airflow": { "command": "python3", "args": ["{path_to_your_folder}/mcp-airflow/server.py"] } } }

- Verify Airflow is running and accessible at the configured URL
- Check authentication credentials (JWT token or username/password)
- Ensure the Airflow REST API is enabled

- Confirm JWT token is valid and not expired
- Verify username and password are correct
- Check that the user has necessary permissions in Airflow

- Ensure DAG IDs and run IDs are correct
- Check that the requested resources exist in Airflow
- Review Airflow logs for additional context

This MCP server uses theAirflow REST API v2. For detailed API documentation, see:

- Airflow REST API Documentation
- Local OpenAPI spec:openapi.json

Provides quality control for biological sequence data using the FastQC and MultiQC tools.