AX – Google’s Open Agentic Orchestrator
Points and comments are a snapshot, not live.
Google releases AX, an open-source agentic orchestrator built on Agent Substrate.
AX is a declarative tool to run agentic tasks at scale, using Kubernetes and Agent Substrate. It provides four core primitives: tasks, workspaces, network policies, and models. Features include sandboxing, sub-second resumption of idle agents, and generative workspace setup. Designed for billions of concurrent tasks per cluster, it targets developers and researchers needing reproducible, isolated agent execution. Born from Google DeepMind research, it is licensed under Apache 2.0.
What commenters are saying
Commenters are skeptical due to Google's history of product sunsets and AI product quality issues, citing poor Gemini integration in Sheets and Docs. Some distinguish open-source projects (Kubernetes, Go) as more stable than consumer products. The thread also notes AX requires Kubernetes and YAML configuration, and compares it to Google's Scion project, which is seen as more integrative. A few criticize Google's security patch practices for Android.