goose is a native, open-source AI agent built in Rust, positioned to run on the user's own machine. It offers three forms of use: a desktop application for macOS, Linux, and Windows; a full CLI for terminal workflows; and an API that can be embedded anywhere. Although it is often used for code-related tasks, the official emphasis is that goose is a general-purpose agent, not limited to code—it can equally be used for research, writing, automation, data analysis, or anything else that needs to get done.
In terms of main capabilities, goose can connect to more than 70 extensions through the Model Context Protocol (MCP), an open standard, including databases, APIs, browsers, GitHub, Google Drive, and more, and it supports adding skills or building extensions yourself. It supports more than 15 model providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, Bedrock, and others; it can use either API keys or, through ACP, existing Claude, ChatGPT, or Gemini subscriptions. goose allows workflows to be captured as portable YAML configurations, making them easy to share with teams, run in CI, and include instructions, extensions, parameters, and sub-recipes. Extensions can also render interactive UI directly within goose Desktop—buttons, forms, visualizations—which is seen as a new way to build agent-driven tools. In addition, goose can spawn independent sub-agents to handle tasks in parallel, such as code review, research, and file processing, thereby keeping the main conversation clean.
On security and standards, goose provides prompt injection detection, tool permission controls, sandbox mode, and an adversary reviewer that monitors unsafe operations. It is built on open standards: MCP is the open standard for connecting AI agents with tools and data sources, and goose is one of the earliest adopters and one of the most deeply integrated projects in the ecosystem, with more than 70 documented extensions; ACP is the standard for communicating with coding agents, and goose can act as an ACP server for Zed, JetBrains, and VS Code to connect to, and can also use ACP agents such as Claude Code and Codex as providers. The project belongs to the Linux Foundation's Agentic AI Foundation, to ensure vendor neutrality, community governance, and long-term openness.
goose has an active community of developers, contributors, and users who work together to build extensions, share recipes, and push the boundaries of local AI agents. Community channels can be used to ask questions, share results, and get help, and participation in open-source development through starring, forking, submitting issues, and contributing code is also welcome. The official project also offers browsing of community-built MCP extensions, as well as resources such as tutorials, in-depth articles, release notes, and community spotlights. goose is available on macOS, Linux, and Windows. Installation commands, the specific extension list, YAML configuration syntax, API details, and charging and pricing information are not currently provided in the available materials.
