GitHub - aaif-goose/goose: an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM

an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM - aaif-goose/goose

Summary

goose is an open-source AI agent built in Rust that runs locally on macOS, Linux, and Windows. It ships as a native desktop app, a CLI, and an embeddable API, and is designed to install, execute, edit, and test code with any supported LLM. The project is part of the Agentic AI Foundation at the Linux Foundation and is released under the Apache-2.0 license.

What it is useful for

goose is positioned as a general-purpose local agent rather than only a code-completion tool. The README states it can be used for research, writing, automation, data analysis, and engineering work. It supports 15+ LLM providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock, and can also use existing Claude, ChatGPT, or Gemini subscriptions via ACP. Through the Model Context Protocol (MCP), it connects to 70+ extensions, which is how engineers extend what the agent can reach on their system. Engineers who want a single agent that can switch between providers and plug into external tools without rewriting their workflow will find this useful.

How engineers can use it

The README documents two entry points. The first is a native desktop app available for macOS, Linux, and Windows, linked from the installation guide. The second is a documented one-liner that installs the CLI:

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

After installing, the README points to a Quickstart and Tutorials on the project documentation site. For teams that want to ship a preconfigured build, the repository includes a "Custom Distributions" guide describing how to package providers, extensions, and branding together. Diagnostics, reporting, and known-issue guidance are linked separately for troubleshooting.

A few details are not covered in the README itself: provider-by-provider configuration, MCP extension setup, and ACP subscription wiring live on the linked docs site, so engineers should follow the Quickstart link in the README for those specifics. The README also does not state performance characteristics or hardware requirements, and it does not enumerate the full provider list, so those would need to be confirmed in the official documentation before adoption.