Speaking · Apache Airflow Town Hall

Bring Your Own AI Agent to Apache Airflow.

A practical look at using an Airflow DAG as the orchestration layer for managed AI agent runtimes across AWS, Google Cloud and Microsoft Azure.

August 2026 · Official monthly town hall

Watch the presentation.

Open the original segment on YouTube

The operating model

One workflow layer, three agent clouds.

Managed agent runtimes give teams a place to deploy AI agents, but production systems still need scheduling, retries, monitoring and lifecycle control. The presentation shows how Apache Airflow can provide that operational layer without coupling a DAG to a single agent framework.

The implementation is grounded in native provider contributions for Amazon Bedrock AgentCore Runtime, Google Cloud Vertex AI Agent Engine and Microsoft Azure AI Foundry Agents. Each integration exposes lifecycle operations as explicit Airflow tasks.

That makes an agent one more governed component in a data workflow: something a team can invoke, update, observe and replace while the surrounding orchestration stays understandable.

Review the Apache Airflow contributions Explore the working demo

Key ideas

What the talk connects.

01

Lifecycle orchestration

Create, invoke, update and delete agent runtimes as explicit DAG operations.

02

Cloud portability

Keep orchestration decisions visible while using the managed runtime that fits each workload.

03

Operational governance

Bring scheduling, retries and observability to agents through the same workflow control plane.