MCP-Airflow-API is an MCP server that exposes Apache Airflow REST API operations as natural-language tools. It supports two runtime modes and dynamic version selection via AIRFLOW_API_VERSION, allowing a single server to work with Airflow v1 (2.x) and v2 (3.0+) REST APIs. The server loads a core set of 43 tools for v1 compatibility and, when operating in v2, adds two asset-management tools for data-aware scheduling (total 45 tools). This enables natural-language control over DAG management, task monitoring, pools, variables, connections, configuration queries, and event logs, with full coverage across API versions. It emphasizes high-scale environments (1000+ DAGs) via smart pagination and advanced filtering, and supports multiple Airflow clusters via environment-configured endpoints. The architecture supports two transport modes: stdio for local MCP clients and streamable-http for Docker/distributed deployments, and can secure remote access with Bearer token authentication in streamable-http mode. Quickstart includes a Docker Compose-based demo with an OpenWebUI and API docs, and there are clear examples for multi-cluster setups across v1 and v2.