Open source · Apache Airflow

Apache Airflow contributions for AI agents.

I contribute provider capabilities to Apache Airflow, with an emphasis on AI agent lifecycle orchestration, Kubernetes operations and the details that make integrations reliable in real DAGs.

13Pull requests authored
9Merged upstream
3Agent clouds covered

Public GitHub snapshot · August 9, 2026

Now in review

Extending where Airflow can operate.

Selected impact

One orchestration layer, three agent clouds.

AWS

Bedrock AgentCore Runtime

Create · Invoke · DeletePR #67984

Google Cloud

Vertex AI Agent Engine

Create · Get · Query · Update · DeletePR #68479

Microsoft Azure

AI Foundry Agents

Create · Update · Run · DeletePR #68799

These integrations let a DAG manage a cloud agent runtime as an explicit lifecycle: provision it, invoke or query it, update it when supported and clean it up. The Agentic Airflow Demo exercises the three paths through the same troubleshooting workflow. Beyond agent runtimes, the merged DmsModifyTaskOperator and DmsReloadTablesOperator make task modification and selective table reload first-class Airflow operations. The reload path includes synchronous and deferrable waits and an end-to-end AWS system test.

Contribution record

The public trail.