Skip to main content
Back

Astronomer Ships Open-Source Agents Giving AI Tools Deep Airflow Knowledge

Astronomer open-sources an MCP server, CLI, and 20+ agent skills that let Claude Code and Cursor write, debug, and deploy Airflow DAGs autonomously

Repo, Full Access
  • Astronomer open-sourced astronomer/agents, an AI agent toolkit for Airflow with 22+ skills, an MCP server, and a CLI tool.
  • Works with 25+ AI coding tools including Claude Code, Cursor, VS Code, Windsurf, and Cline via a one-command install.
  • The MCP server connects AI agents directly to any Airflow 2.x/3.x REST API for DAG management, triggering, task logs, and health monitoring.
  • Skills cover the full pipeline lifecycle: DAG authoring, debugging, lineage tracing, warehouse analysis, dbt integration, and Airflow 2-to-3 migration.
  • Together AI used this stack in production, having agents author 700+ dbt models and cutting pipeline development from weeks to hours.
  • Free and Apache 2.0 licensed; works with open-source Airflow, no Astronomer managed platform required.

AI coding agents are surprisingly bad at writing correct Airflow code. They know Python, but they don't know Airflow's operator quirks, the difference between 2.x and 3.x patterns, or how to trace a pipeline failure through task logs. Astronomer's new open-source agents repo is a direct fix for that: a collection of agent skills, an MCP server, and a CLI that give any AI coding tool deep, production-tested Airflow knowledge.

The toolkit ships 22+ specialized skills spanning the full data pipeline lifecycle, from authoring DAGs with best practices to debugging pipeline failures, tracing data lineage, profiling warehouse tables, and managing local and cloud deployments. It also includes an Airflow MCP server providing full REST API integration for DAG management, triggering, task logs, and system health monitoring.

Three things in one repo

The repo includes an MCP server for Airflow, a CLI tool (af) for interacting with Airflow from your terminal, and skills that extend AI coding agents with specialized capabilities for working with Airflow and data warehouses. Here's what each piece does:

  • Skills are modular markdown-based instruction sets that AI agents invoke automatically based on what you ask. They encode Airflow best practices, version-specific patterns, and debugging strategies that a general-purpose LLM wouldn't know.
  • The MCP server (astro-airflow-mcp) connects your AI tool directly to a live Airflow instance via the REST API. It's a Model Context Protocol server for Apache Airflow built with FastMCP. The MCP auto-discovers your project when you run Claude Code from an Airflow project directory containing