The GitHub Actions job "npm_and_yarn in /airflow-core/src/airflow/api_fastapi/auth/managers/simple/ui - Update #1270009044" on airflow.git/main has succeeded. Run started by GitHub user dependabot[bot] (triggered by dependabot[bot]).
Head commit for run: 86e63e8a087e160118168582564f0b6da9fea1bc / Kaxil Naik <[email protected]> AIP-99: Add MCPToolset and MCPHook for MCP server integration (#62904) Add support for connecting to MCP (Model Context Protocol) servers from Airflow AI agents. MCPToolset resolves server config from an Airflow connection and delegates to PydanticAI's MCP server classes, giving users connection management and secret backend integration. - MCPHook: dedicated `mcp` connection type with UI fields for transport (http/sse/stdio), command, args, and auth token - MCPToolset: thin wrapper implementing AbstractToolset that reads MCP server config from an Airflow connection - Auth token from connection password passed as Bearer header - Lifecycle delegation via __aenter__/__aexit__ to keep MCP connections open across tool calls - Example DAGs showing MCPToolset with connections, multiple servers, and direct PydanticAI MCP server usage - Documentation covering three tiers: Airflow-native toolsets, direct PydanticAI MCP servers, and any AbstractToolset - MCPToolset delegates to MCPHook instead of duplicating connection logic - Add tool_prefix parameter to MCPHook - Fix hook tests to patch pydantic_ai.mcp (lazy import inside get_conn) - Add spelling words: Streamable, hardcoding, stdin, tradeoff Report URL: https://github.com/apache/airflow/actions/runs/22727716436 With regards, GitHub Actions via GitBox --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
