MCP Server Apache Airflow

What yangkyeongmo@/mcp-server-apache-airflow MCP server does

yangkyeongmo@/mcp-server-apache-airflow provides an MCP interface over Apache Airflow's REST API. It is intended for MCP clients that need to work with Airflow objects through standardized tool calls rather than constructing HTTP requests themselves. The implementation uses Apache Airflow's official client library.

The supported resource areas include DAG definitions, DAG runs, task metadata and instances, Airflow variables, connections, pools, and XCom entries. The feature table also lists dataset operations, including listing datasets, although the supplied excerpt ends before the complete dataset section is shown.

How it works

The server translates MCP requests into operations against Airflow REST API endpoints under /api/v1. Read operations can retrieve DAGs, run details, task information, logs, variables, connections, pools, XCom entries, and related resources. Write operations cover actions such as pausing or unpausing DAGs, creating or updating DAG runs, changing task states, managing variables and connections, and modifying pools.

The API mapping follows Airflow resource paths. For example, DAG operations use /api/v1/dags, DAG-run operations use paths beneath a DAG identifier, and task-instance operations use paths beneath a specific run and task. The implementation also includes operations for clearing runs or task instances, reparsing a DAG source, testing a connection, and setting a DAG-run note.

Tools and capabilities

The documented capabilities cover:

• Listing, retrieving, updating, pausing, unpausing, deleting, and batch-patching DAGs.
• Reading DAG source files and requesting DAG-file reparsing.
• Listing, creating, updating, deleting, clearing, and inspecting DAG runs.
• Reading task definitions, task instances, task tries, and task-instance logs.
• Updating task instances and changing the state of multiple task instances.
• Creating, reading, updating, and deleting Airflow variables.
• Creating, reading, updating, deleting, and testing Airflow connections.
• Creating, reading, updating, and deleting pools.
• Listing and retrieving XCom entries for a task instance.

These operations include both monitoring-oriented access and mutating actions. Agents using the server may therefore need permissions appropriate to the Airflow actions they are allowed to perform.

Limitations and notes

yangkyeongmo@/mcp-server-apache-airflow is described as an Airflow REST API wrapper, so its usable scope depends on the API and permissions available from the target Airflow deployment. The material does not specify an Airflow version matrix, authentication method, required environment variables, transport configuration, or a command for launching the package.

The feature table marks the listed endpoints as implemented, but it does not document request schemas, response formats, pagination behavior, error handling, or permission requirements for individual operations. Review the target Airflow deployment and the project's current source before enabling write operations in an automated agent workflow.