Google has open-sourced Agent Executor and Agent Substrate to help enterprises deploy resilient, secure, and scalable AI agents across Kubernetes and distributed production environments.
Google has introduced Agent Executor, an open-source runtime standard designed for executing, resuming, and deploying long-running AI agents across distributed production environments. Alongside it, the company also unveiled Agent Substrate, an open-source orchestration layer built on Kubernetes for hyperscale AI agent workloads.
The runtime is designed for AI workflows that may run for hours or days while surviving outages, interruptions, client disconnections, and human-in-the-loop approval stages. Agent Executor uses event logging and snapshotting to enable durable execution and workflow recovery for agents, tools, skills, sandboxes, and agent harnesses.
The platform also supports connection recovery, allowing clients to reconnect and resume workflows from the last received sequence. Another feature, trajectory branching, enables developers to test multiple workflow paths from checkpoints while preserving workflow context and state.
Security is handled through sandboxed isolation designed to reduce harmful side effects and protect against malicious activity from untrusted LLM-generated code. Google said the runtime supports Google Kubernetes Engine Sandbox and Kata Containers, while enforcing a default-deny network security posture.
Agent Executor works with LangChain, LangGraph, Gemini API Managed Agents, and Google’s Agent Development Kit. It can also run on self-managed infrastructure.
Agent Substrate adds Kubernetes-native scheduling, rapid scaling, Pod Snapshot integration, and support for hundreds of millions of registered agents handling millions of short AI tool calls.
