What features does Airflow MCP server provide?

Q: What features does Airflow MCP server provide?

DAG Management: Expose endpoints to list DAGs, get details, pause/unpause, update, delete, access DAG sources, patch multiple DAGs, and reparse DAG files.. DAG Runs: Lifecycle management for DAG runs: list, create, read, update, delete; batch retrieval; clear; set notes.. Tasks: Task-level operations: list tasks, get task details and task instances, access logs, update state, clear, and manage task tries.. Variables: CRUD operations for Airflow Variables: list, create, get, update, delete.. Connections: Manage Airflow Connections: list, create, get, update, delete, and test connections.. Datasets: Dataset-related endpoints: list, get, dataset events, create dataset events, and manage DAG dataset queued events.. Monitoring: Health monitoring: Get Health endpoint to observe server and Airflow status.. Config: Get Airflow configuration via the MCP API.

Airflow MCP Server 2025: Features & Setup Guide

Overview

This MCP server implements the Model Context Protocol to expose Apache Airflow resources through a unified MCP interface. It wraps Airflow's REST API using the official Python client library, ensuring compatibility and maintainability, and maps MCP operations to a broad set of Airflow capabilities. The server exposes endpoints for DAG management (listing, detailing, pausing, unpausing, updating, deleting; accessing DAG sources; reparse of DAG files), DAG runs (list, create, read, update, delete; batch retrieval; clear; set notes), tasks (listing and detailing tasks, task instances, logs, updates to states, clearing, and retry/tries), variables, connections, pools, XComs, and datasets, along with monitoring (health check) and configuration endpoints. It supports API group selection via --apis, and can run in read-only mode to restrict destructive operations. Authentication can be Basic or JWT, with JWT preferred if both are provided. Deployment examples include Claude Desktop and Smithery configurations, and the server can run via uvx/uv or direct Python execution. Tests and CI are provided using pytest and GitHub Actions.

Features

DAG Management

Expose endpoints to list DAGs, get details, pause/unpause, update, delete, access DAG sources, patch multiple DAGs, and reparse DAG files.

DAG Runs

Lifecycle management for DAG runs: list, create, read, update, delete; batch retrieval; clear; set notes.

Tasks

Task-level operations: list tasks, get task details and task instances, access logs, update state, clear, and manage task tries.

Variables

CRUD operations for Airflow Variables: list, create, get, update, delete.

Connections

Manage Airflow Connections: list, create, get, update, delete, and test connections.

Datasets

Dataset-related endpoints: list, get, dataset events, create dataset events, and manage DAG dataset queued events.

Monitoring

Health monitoring: Get Health endpoint to observe server and Airflow status.

Config

Get Airflow configuration via the MCP API.