AX – Google’s Open Agentic Orchestrator
AX is presented as an open agentic orchestrator under the name "AX – Google's Open Agentic Orchestrator," linked from Hacker News to agentexecutor.io. The page's pitch is that a user declares an agentic task and AX runs it at scale: it sandboxes the task, wires up its workspace, fences its network, and, per the page, helps run billions of them per cluster. Users can either use a single task per agent or compose as many as the agent needs. The page demonstrates the workflow through a terminal session. A task.yaml file defines two resources with apiVersion ax.io/v1alpha1: a Workspace named golang with a git repo pointing at https://github.com/golang/go.git on branch "my-fix", and a Task named test referencing that workspace with the goal "Ensure that Go tool chain is available and is built from source" and debug set to true. Running ax apply -f task.yaml creates workspace.ax.io/golang and task.ax.io/test. The ax watch task test command shows the task moving from Phase: Pending at 10:42:01 to Phase: Running at 10:42:05 with WorkerIP 10.20.3.67, and ax get tasks lists test in namespace default in phase Running. The session then uses ax ssh to list /workspace, build the Go source, inspect processes (showing /usr/local/bin/ax-task-runner and a go build process), and create notes.txt, followed by ax suspend, ax resume, and ax delete task test.
Developers can define agentic tasks in YAML and control them through CLI commands like apply, watch, ssh, suspend, resume, and delete.