What’s new here

Goose ships a native desktop app, a full CLI, and an embeddable API from the same Rust codebase. It supports the Agent Communication Protocol (ACP) for provider subscriptions, so you can use an existing Claude, ChatGPT, or Gemini subscription rather than raw API keys. It connects to 70+ extensions via the Model Context Protocol (MCP) open standard.

What it does

Goose is a general-purpose AI agent that runs on your local machine. The GitHub description says it can install, execute, edit, and test code. The README describes research, writing, automation, and data analysis as valid uses beyond code.

It connects to 15+ LLM providers including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock. It speaks the Model Context Protocol and lists 70+ extensions available through that standard.

The project lives under the Linux Foundation via the Agentic AI Foundation (AAIF). A CUSTOM_DISTROS.md documents how to build your own distribution with preconfigured providers, extensions, and branding.

Who it’s for

Developers who want an agent they can run locally from the terminal or a desktop app. Also useful for teams that need to embed agent capability in a larger tool via the API, or that want to ship a branded internal distribution with locked-down provider and extension defaults.

Try it

Prebuilt releases are available for macOS (arm64), Linux (arm64 and x86_64, glibc and musl, including Vulkan variants), and Windows. Download from the releases page or install the CLI with:

curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash

For the desktop app, see the installation docs. The quickstart covers first-run provider setup.

How mature is it

55,395 stars, 6,457 forks, 100+ contributors (capped), 100+ releases (capped), latest v1.55.0 tagged 2026-10-16, last push 2026-10-18, 100+ commits in the last 90 days (capped), 320 open issues, 178 open PRs. Apache-2.0 license. Part of the Linux Foundation via AAIF. Active by every measure.