Googlefeature
May 20, 2026·Source

Agent Executor

Google's open-source runtime for deploying distributed AI agents that can pause, resume, and run on your own infrastructure.

Agent Executor announcement image
  • Agents resume mid-task after outages by replaying from a saved event log and snapshot
  • Execution is sandboxed so each component runs in isolation and cannot cause harmful side effects across the system
  • Supports LangChain/LangGraph, Google's Agent Development Kit, and the Agent2Agent Protocol
  • Available now in preview; open-source at github.com/google/ax
  • Works alongside Agent Substrate, a new open-source Kubernetes abstraction also released today for running millions of agents at scale

Best for

Teams building agent workflows that run for hours or days and need to survive outages or infrastructure changes

When to use it

Deploying agents to custom or on-premises infrastructure, or when you need durable execution that resumes from a checkpoint

Who benefits

Platform engineers and enterprise teams building production AI agent systems who want to avoid vendor lock-in

Join the newsletter to get ahead with AI.

I share strategies, new modules, free guides, and tools to get you better at using AI in your work.

Leading a team? See team programs →