MCP server and CLI letting AI agents reverse engineer native binaries, Electron apps, .NET assemblies and websites.
google/ax
AX is an open-source declarative orchestrator for running large numbers of sandboxed autonomous agent tasks in a Kubernetes cluster.
- Stars
- 13,963
- Forks
- 723
- Open issues
- 91
- Last commit
- 14 hours ago
What is it?
AX lets you describe an agentic task in YAML using three resource kinds: Task, Workspace, and Model. A Task runs in an isolated sandbox with CPU and memory limits, a Workspace pre-wires Git repositories, MCP servers, and skill packages so agents start with their tools in place, and a Model configures which LLM the platform uses, with credentials taken from a Kubernetes secret. The manifests use the ax.io/v1alpha1 API and are applied with the ax CLI, which follows kubectl-style verbs.
AX runs on top of Agent Substrate for sandboxed execution. It supports suspending an idle task with a checkpoint and resuming it later, and shelling into a running sandbox with ax ssh for debugging. The CLI talks to the control plane over gRPC and resolves the cluster from the active kubectl context.
The project is in heavy development and its concepts, protocols and APIs may change with breaking updates before a stable release. Its installation requires a Kubernetes cluster with Agent Substrate already running.
Key features
- 01Declarative Task, Workspace and Model manifests applied with a single command.
- 02Sandboxed task execution with CPU and memory limits.
- 03Suspend and resume of running agent tasks via checkpointing.
- 04Interactive shell access into a running sandbox with ax ssh.
- 05kubectl-style CLI that follows the active Kubernetes context.
Use cases
- Running many untrusted or autonomous agent workloads isolated inside a Kubernetes cluster.
- Pre-configuring Git repositories and MCP tool servers so agents start with the needed context.
- Pausing long-running idle agents and resuming them later without losing state.
Topics
FAQ
- What is ax?
- AX is an open-source declarative orchestrator for running large numbers of sandboxed autonomous agent tasks in a Kubernetes cluster. AX lets you describe an agentic task in YAML using three resource kinds: Task, Workspace, and Model. A Task runs in an isolated sandbox with CPU and memory limits, a Workspace pre-wires Git repositories, MCP servers, and skill packages so agents start with their tools in place, and a Model configures which LLM the platform uses, with credentials taken from a Kubernetes secret. The manifests use the ax.io/v1alpha1 API and are applied with the ax CLI, which follows kubectl-style verbs.
- Is ax free to use? What license does it have?
- ax is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is ax still maintained?
- Actively maintained. The last commit to ax was on Oct 18, 2026.
- What can I use ax for?
- Running many untrusted or autonomous agent workloads isolated inside a Kubernetes cluster. Pre-configuring Git repositories and MCP tool servers so agents start with the needed context. Pausing long-running idle agents and resuming them later without losing state.
- Where can I get ax?
- The source code and installation instructions are on GitHub: https://github.com/google/ax
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A toolkit that generates agent-native command-line interfaces for existing software, plus a CLI-Hub registry for installing ready-made ones.
A proxy and CLI that gives AI agents one base URL and token to call a curated catalog of paid third-party APIs, plus your team's own keys and tools.
Zero-trust sandboxed runtime on Kubernetes for running large numbers of untrusted, mostly-idle AI agents with snapshot-based suspend and resume.
Self-hosted infrastructure that runs untrusted code in lightweight VM sandboxes, managed through a CLI, HTTP API and Python SDK.
A plugin-based harness that gives Claude Code, Codex and other coding agents shared agents, skills, hooks, memory and security scanning.