Goose

goose is a general-purpose AI agent that runs on your machine, for coding as well as research, writing, automation, and data analysis. It is written in Rust and comes as a native desktop app for macOS, Linux, and Windows, a CLI for terminal workflows, and an API for embedding it in other software. The project is hosted by the Agentic AI Foundation (AAIF) at the Linux Foundation.

goose works with 15+ model providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock, using API keys or existing Claude, ChatGPT, or Gemini subscriptions through the Agent Client Protocol (ACP). Capabilities come from 70+ extensions connected over the Model Context Protocol (MCP), covering databases, APIs, browsers, GitHub, Google Drive, and more, and from skills you add or extensions you build.

Workflows can be captured as recipes: portable YAML configs that carry instructions, extensions, parameters, and subrecipes, can be shared with a team, and can run in CI. MCP Apps let extensions render interactive interfaces inside the desktop app, and custom distributions package goose with preconfigured providers, extensions, and branding.

Features

• Three surfaces: a native desktop app, a full CLI, and an embeddable API
• Model providers: 15+ including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock
• Subscription access: existing Claude, ChatGPT, or Gemini subscriptions usable through ACP
• MCP extensions: 70+ extensions for databases, APIs, browsers, GitHub, Google Drive, and more
• Recipes: YAML workflow definitions with instructions, extensions, parameters, and subrecipes, runnable in CI
• MCP Apps: extensions can render interactive UI inside goose Desktop
• Custom distributions: build a goose distro with preconfigured providers, extensions, and branding