MCP Server for Apache Airflow

Instructions

Implementation Reference

• src/airflow/dagrun.py:174-185 (handler)

  The core handler function for the 'get_dag_run' tool. Fetches the DAG run details from Airflow API, adds a UI URL using the helper, converts to dict, and returns as MCP TextContent.

  async def get_dag_run(
      dag_id: str, dag_run_id: str
  ) -> List[Union[types.TextContent, types.ImageContent, types.EmbeddedResource]]:
      response = dag_run_api.get_dag_run(dag_id=dag_id, dag_run_id=dag_run_id)
  
      # Convert response to dictionary for easier manipulation
      response_dict = response.to_dict()
  
      # Add UI link to DAG run
      response_dict["ui_url"] = get_dag_run_url(dag_id, dag_run_id)
  
      return [types.TextContent(type="text", text=str(response_dict))]

• src/airflow/dagrun.py:17-29 (registration)

  Module-level registration function that includes the 'get_dag_run' tool in its list of tools to be registered with the MCP server.

  def get_all_functions() -> list[tuple[Callable, str, str, bool]]:
      """Return list of (function, name, description, is_read_only) tuples for registration."""
      return [
          (post_dag_run, "post_dag_run", "Trigger a DAG by ID", False),
          (get_dag_runs, "get_dag_runs", "Get DAG runs by ID", True),
          (get_dag_runs_batch, "get_dag_runs_batch", "List DAG runs (batch)", True),
          (get_dag_run, "get_dag_run", "Get a DAG run by DAG ID and DAG run ID", True),
          (update_dag_run_state, "update_dag_run_state", "Update a DAG run state by DAG ID and DAG run ID", False),
          (delete_dag_run, "delete_dag_run", "Delete a DAG run by DAG ID and DAG run ID", False),
          (clear_dag_run, "clear_dag_run", "Clear a DAG run", False),
          (set_dag_run_note, "set_dag_run_note", "Update the DagRun note", False),
          (get_upstream_dataset_events, "get_upstream_dataset_events", "Get dataset events for a DAG run", True),
      ]

• src/main.py:6-20 (registration)

  Imports the get_all_functions from dagrun module (line 9) and maps it in APITYPE_TO_FUNCTIONS dict (line 28, not shown) for top-level tool registration.

  from src.airflow.config import get_all_functions as get_config_functions
  from src.airflow.connection import get_all_functions as get_connection_functions
  from src.airflow.dag import get_all_functions as get_dag_functions
  from src.airflow.dagrun import get_all_functions as get_dagrun_functions
  from src.airflow.dagstats import get_all_functions as get_dagstats_functions
  from src.airflow.dataset import get_all_functions as get_dataset_functions
  from src.airflow.eventlog import get_all_functions as get_eventlog_functions
  from src.airflow.importerror import get_all_functions as get_importerror_functions
  from src.airflow.monitoring import get_all_functions as get_monitoring_functions
  from src.airflow.plugin import get_all_functions as get_plugin_functions
  from src.airflow.pool import get_all_functions as get_pool_functions
  from src.airflow.provider import get_all_functions as get_provider_functions
  from src.airflow.taskinstance import get_all_functions as get_taskinstance_functions
  from src.airflow.variable import get_all_functions as get_variable_functions
  from src.airflow.xcom import get_all_functions as get_xcom_functions

• src/airflow/dagrun.py:32-33 (helper)

  Helper function used by the handler to generate the UI URL for the DAG run.

  def get_dag_run_url(dag_id: str, dag_run_id: str) -> str:
      return f"{AIRFLOW_HOST}/dags/{dag_id}/grid?dag_run_id={dag_run_id}"

• src/main.py:95-96 (registration)

  The loop where all collected functions, including get_dag_run, are registered as MCP tools using app.add_tool.

  for func, name, description, *_ in functions:
      app.add_tool(Tool.from_function(func, name=name, description=description))