Google AX

Google AX is an open-source agent harness and execution orchestrator developed by Google. It provides runtime infrastructure and execution sandboxing designed to bridge terminal-based tool environments with Google's Antigravity agent architecture and Gemini/Vertex AI model backends. Day-to-day, AX operates as a command-line harness (`cmd/ax`) interfacing with a local sidecar service. The system manages conversational state across interactions, maintaining persistent execution directories (`AX_HARNESS_WORKDIR`) and routing tool executions such as shell commands, code generation, and repository edits. It communicates with language models primarily through Google Cloud Vertex AI and the Google GenAI SDK ecosystem. What distinguishes AX from conventional standalone CLI agents is its decoupled harness architecture. Instead of tightly coupling model calls with the user terminal, AX implements a client-sidecar model that enforces workspace boundaries, isolates execution environments (including support for sandboxed/gVisor environments), and cleanly serializes conversation turns to durable storage for session resumption. Google AX is open-source software distributed under an Apache 2.0 license. Model inference costs depend on your underlying Google Cloud Vertex AI or Google AI Studio billing configuration.

Tags: agent harness, AI, ai agent, cli, developer tools, OpenSource

Visit Google AX

Key features

  • Harness-sidecar architecture — Decouples the command-line interface from the agent execution runtime using a local sidecar server over WebSocket connections. - Authoritative workspace directory isolation — Enforces a strict workspace boundary (AX_HARNESS_WORKDIR) so command executions and relative path resolutions do not drift into the ambient process directory. - Vertex AI native integration — Connects directly to Google's model infrastructure using standard environment variables (GOOGLE_GENAI_USE_VERTEXAI, GOOGLE_CLOUD_PROJECT, and GOOGLE_CLOUD_LOCATION). - Durable session persistence — Manages conversation turns in segmented state directories, enabling safe resumption and state retention across runs. - Built-in diagnostics — Includes an ax doctor command to inspect, validate, and troubleshoot local sidecar connections, runtime dependencies, and asset integrity.

Use cases

Use Case: Sandboxed Workspace Execution Developers running agentic code modifications in restricted execution containers or gVisor-based environments can deploy AX to guarantee that tool calls execute strictly within designated workspace roots without leaking ambient process paths.

Use Case: Persistent Multi-Turn Task Orchestration Engineers tackling long-running refactoring or debugging tasks can leverage AX's session state directory persistence to interrupt, inspect, and resume agent workflows without losing turn context.

Use Case: Vertex AI Enterprise Agent Harnessing Teams building internal developer tooling on top of Google Cloud's enterprise model endpoints can use AX as a structured CLI execution harness that respects standard cloud credentialing and project configurations.

Who it is for

Google AX is built for systems engineers, AI tooling developers, and enterprise platform teams implementing sandboxed, repeatable agent execution workflows on top of Google Cloud Vertex AI. It is less suited for end users looking for a plug-and-play, single-binary interactive chat UI without infrastructure setup.

Install guide

Installation Google AX is written in Go with supporting Python sidecar components. Clone the repository and build the CLI binary: bash git clone https://github.com/google/ax.git cd ax go build -o ax ./cmd/ax Ensure Python and required runtime dependencies (google-antigravity) are installed in your environment.

First Run Configure your Google Cloud credentials and environment variables before starting the harness: bash export GOOGLE_GENAI_USE_VERTEXAI=true export GOOGLE_CLOUD_PROJECT="your-gcp-project-id" export GOOGLE_CLOUD_LOCATION="us-central1" export AX_HARNESS_WORKDIR="$(pwd)" # Run the diagnostic tool to verify environment and sidecar health ./ax doctor

Example Session Run the agent harness against your current project directory: bash ./ax run "Audit the repository for broken unit tests and apply necessary fixes"

Part of CLI Coding Agents

Categories: AI Coding Agents, Chatbot Development, Developer Tools

Related CLI agents

  • ActPlane — ActPlane is an open-source agent execution and policy enforcement harness developed by the Eunomia-bpf organization. It functions as a security…
  • hol-guard — hol-guard is a security scanner and agent guardrail harness developed by Hashgraph Online. Built to safeguard agentic runtime environments and…
  • Skillreaper — Skillreaper is an open-source skill lifecycle management and auditing tool for AI agent ecosystems, developed by thousandflowers. It analyzes agent…
  • Agent Island — Agent Island is an open-source agent orchestration harness created by developer tristan666666 on GitHub. Designed as a framework for running…
  • Stratless — Stratless is an open-source agent harness and execution orchestrator developed by the Stratless AI team. It provides a modular infrastructure…
  • NemoClaw — NemoClaw is an open-source agent orchestration harness and execution framework developed by NVIDIA. It provides structured runtimes, modular skill dispatching,…
  • AI Tools
  • Categories
  • Industries
  • CLI Coding Agents
  • MCP Servers
  • MCP Categories