Connect Airflow to AI: Let Claude and Cursor Check DAGs, Task Runs, and Logs

Why Connect Airflow to AI?

Airflow is where data teams orchestrate pipelines: DAGs define dependencies, tasks execute transforms, logs capture what happened. When you need to check if a DAG succeeded, read a task's logs, or see what's scheduled next, you open the Airflow UI, navigate through the tree view, read the output, switch back.

With an MCP server pointed at Airflow, AI agents can query your orchestration layer directly. Check DAG runs, read task logs, browse schedules without leaving your terminal.

What AI Agents Can Do With Airflow Access

Once connected, Claude, Cursor, or Codex can:

• Check DAGs — "Did the daily_etl DAG succeed this morning?"
• Read logs — "What error did the transform_orders task throw?"
• Browse schedules — "What DAGs are scheduled to run in the next hour?"
• Check dependencies — "What tasks does the load_warehouse task depend on?"
• Monitor state — "Are there any DAGs in a failed state right now?"
• List runs — "Show me the last 5 runs for the analytics_pipeline DAG"

How It Works

1. Open DataFaucet and enter your Airflow web UI URL
2. Browse Airflow — click through DAGs, task instances, logs, schedules
3. DataFaucet captures the read operations as MCP tools
4. Deploy — your AI agents now have structured access to query Airflow

You're browsing the Airflow web interface, not sharing connection strings. DataFaucet captures HTTP interactions from the web UI. Your Airflow credentials stay in your browser session.

Example Queries

"Did the daily_etl DAG succeed last night?"
"What error caused the transform_users task to fail?"
"What DAGs are currently running?"
"When is the next scheduled run for analytics_pipeline?"
"Show me failed tasks from the last 24 hours"

When This Saves Time

• Pipeline debugging — read task logs without opening Airflow
• Morning checks — verify overnight DAGs ran successfully
• Incident response — quickly find which task failed and why
• Code review — verify DAG dependencies before approving changes
• Planning — check schedule density during capacity discussions

Security Model

DataFaucet connects through the Airflow web UI. Access is scoped to whatever your Airflow RBAC permissions allow. If you only have read access to certain DAGs, the MCP server has the same constraints. No connection strings or Fernet keys stored separately.

Getting Started

Open Airflow in your browser, copy the web UI URL. Paste it into DataFaucet, browse through DAGs, task instances, and logs. Deploy. Your agents can now query your orchestration layer from your terminal.

Create your Airflow MCP server →