This server provides a curated interface for inspecting and operating Apache Airflow 2 via its REST API, offering both read-only access and guarded write capabilities.

Read capabilities:

• get_airflow_version – Check the Airflow version and confirm API connectivity

• get_airflow_health – View health status of Airflow components (metadatabase, scheduler, triggerer, dag-processor)

• list_dags – List DAGs with pause state, schedule, owners, and tags (filterable by tag, paused state, and dag_id pattern)

• get_dag – Get details of a single DAG by ID

• list_dag_runs – List runs of a DAG, most recent first (filterable by state, limit, offset)

• get_dag_run – Get a single DAG run by DAG ID and run ID

• list_task_instances – List task instances in a DAG run with states and timings (filterable by state)

• get_task_instance – Get a single task instance by DAG ID, run ID, and task ID

• get_task_logs – Read logs for a specific task attempt (supports tailing for large logs)

• list_import_errors – List DAG parse/import errors with filenames and stack traces

• list_pools – List worker pools with slot usage (occupied, running, queued, open)

Write capabilities (disabled when AIRFLOW_MCP_READ_ONLY=true):

• trigger_dag_run – Trigger a new DAG run, optionally with a custom run ID, logical date, config, and note

• set_dag_paused – Pause or unpause a DAG to stop/resume scheduling

• clear_task_instances – Clear (retry) task instances; supports dry_run=True to preview affected tasks without making changes, with options to filter by task IDs, run ID, failed-only, and include upstream/downstream tasks

Allows inspection and operation of Apache Airflow over its REST API, providing tools for managing DAGs, DAG runs, task instances, logs, pools, and import errors.

1. Click on "Install Server".

2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.

3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@airflow-mcp-server what's the health of my Airflow instance?"

That's it! The server will respond to your query, and you can continue using it as needed.

Here is a step-by-step guide with screenshots.