Airflow runs on your own machine — the client starts it, so there is no endpoint to ping. 97 installs a week from npm. Last commit 10 Jul 2026.
Airflow MCP — list DAGs/runs/task instances, tail logs, trigger and clear (write-gated)
Today is the operative word: we check Airflow every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add airflow -e 'AIRFLOW_API_URL=<AIRFLOW_API_URL>' -e 'AIRFLOW_USERNAME=<AIRFLOW_USERNAME>' -e 'AIRFLOW_PASSWORD=<AIRFLOW_PASSWORD>' -- npx -y @us-all/airflow-mcp
{
"mcpServers": {
"airflow": {
"command": "npx",
"args": [
"-y",
"@us-all/airflow-mcp"
],
"env": {
"AIRFLOW_API_URL": "<AIRFLOW_API_URL>",
"AIRFLOW_PASSWORD": "<AIRFLOW_PASSWORD>",
"AIRFLOW_USERNAME": "<AIRFLOW_USERNAME>"
}
}
}
}
[mcp_servers.airflow]
command = "npx"
args = ["-y", "@us-all/airflow-mcp"]
env = { AIRFLOW_API_URL = "<AIRFLOW_API_URL>", AIRFLOW_USERNAME = "<AIRFLOW_USERNAME>", AIRFLOW_PASSWORD = "<AIRFLOW_PASSWORD>" }
{
"mcpServers": {
"airflow": {
"command": "npx",
"args": [
"-y",
"@us-all/airflow-mcp"
],
"env": {
"AIRFLOW_API_URL": "<AIRFLOW_API_URL>",
"AIRFLOW_PASSWORD": "<AIRFLOW_PASSWORD>",
"AIRFLOW_USERNAME": "<AIRFLOW_USERNAME>"
}
}
}
}
{
"servers": {
"airflow": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@us-all/airflow-mcp"
],
"env": {
"AIRFLOW_API_URL": "<AIRFLOW_API_URL>",
"AIRFLOW_PASSWORD": "<AIRFLOW_PASSWORD>",
"AIRFLOW_USERNAME": "<AIRFLOW_USERNAME>"
}
}
}
}
Put your own values in place of
<AIRFLOW_API_URL> (Airflow host base, e.g. 'http://airflow.example.com:8080'. Server prepends /api/v2 internally; trailing /api/v1 or /api/v2 is stripped if supplied), <AIRFLOW_USERNAME> (Airflow username (used to mint a JWT via SimpleAuthManager /auth/token)), <AIRFLOW_PASSWORD> (Airflow password (used to mint a JWT via SimpleAuthManager /auth/token)).
The author’s registry entry marks them as required.
Optional settings from the registry entry, left out of the block above:
AIRFLOW_ALLOW_WRITE (Set to 'true' to enable trigger-dag and clear-task tools. Default read-only), AIRFLOW_TOOLS (Comma-separated category allowlist (airflow,meta)), AIRFLOW_DISABLE (Comma-separated category disablelist).
Observe and operate Hatchet workflows from an AI agent — runs, logs, trigger, cancel, replay.
A committed markdown dev log and task graph that agents read and write over MCP.
CircleCI's MCP server: inspect projects, runs, workflows, jobs, logs, tests, and artifacts.
Create, search, and list Changologs logs and tasks
Local LLM does Claude's bulk work: reads logs and files, researches the web, writes test-gated code.
Local LLM does Claude's bulk work: reads logs and files, researches the web, writes test-gated code.
Run core Docker container operations: list, run, inspect, logs, images and volumes.
Langfuse public API (Cloud and self-hosted): traces, prompts, datasets and gated writes.
Answers built from our own checks of this server.