google/ax
Declarative orchestration runtime for autonomous agent workloads
- Chart
- Not on a chart right now
- Stars
- 14.8k
What it is
Agent Executor (AX) is Google's open agentic orchestration runtime built on Agent Substrate to run autonomous agent workloads in sandboxed environments at cluster scale. It uses Kubernetes-style declarative YAML manifests (Task, Workspace, and Model) to wire up workspaces and manage agent lifecycles.
Who it is for
Platform engineers and developers running and scaling isolated AI agent workloads on Kubernetes clusters.
Why it is trending
With 14,754 total stars and a v0.3.1 release, AX stands out by managing stateful, isolated agent workloads through three declarative primitives and familiar kubectl-style CLI commands.
Main features
- Declarative Task, Workspace, and Model definitions via ax.io/v1alpha1 manifests
- Pre-wired Git repositories, MCP servers, and skill packages in Workspace specs
- Checkpoint and restore agent state with ax suspend and ax resume
- Direct shell access into running sandboxes via ax ssh
Summarized on November 17, 2026 from the README and repository data. It can be wrong; check the source before relying on it. Report a wrong summary