Google AX is an Apache-2.0 orchestrator that runs declarative agent tasks in sandboxed, network-fenced workspaces at scale on Kubernetes.
Yes, Google AX is open source under the Apache-2.0 license.
Google AX is free to use.
Yes, Google AX can be self-hosted (the source is available under the Apache-2.0 license).
Apache-2.0 agentic orchestrator from Google (Go) for running autonomous agent workloads at scale. You declare an agent task in a YAML manifest, its workspace and its gateway, and AX sandboxes the task, wires up the workspace, fences its network, and runs it on Kubernetes through a control plane. It integrates with LLM providers you supply, so it handles isolation and scale-out while you bring the model. Early observers describe it as Kubernetes for agent executions.
Once agents graduate from a demo to doing real work across a team, the problem stops being the prompt and becomes operations: how do you run twenty agent tasks without one of them wandering out onto the network or trampling a shared machine. AX is Google's answer, declare the task, and it runs sandboxed and network-fenced, at scale, on infrastructure you own. For a GTM or RevOps team that wants agents touching real systems with a blast radius you can set, that isolation is the feature. The honest read: it assumes Kubernetes and a control plane, so this is platform work, not a weekend install, and you still bring and pay for the models. As the piece that makes running many agents an operational discipline instead of a hope, it is worth the pilot.