How to Get Started with goose and Where to Get Help

Updated

goose is an open source, general-purpose AI agent that runs on your machine, and you can start using it by downloading the native desktop app for macOS, Linux, or Windows, or by using its CLI or API. The official site provides a Get started entry point for installation, and once you're running, help is available through the project's Discord community, GitHub repository, and a browsable library of community-built MCP extensions.

Getting started

goose ships in three forms, so the right entry point depends on how you want to work:

  • Desktop app — a native application for macOS, Linux, and Windows. Best if you want a graphical interface and interactive extension UIs.
  • CLI — a full command-line interface for terminal workflows. Best if you live in a shell or want to script agent runs.
  • API — an interface for embedding goose into your own tools and services.

The site's Get started section is the place to begin, and a See goose in action section shows what the agent looks like in practice before you commit to installing anything.

goose is built in Rust, which the project cites for performance and portability across these platforms.

What you can use it for

goose is described as general-purpose rather than code-only. The documented use cases include research, writing, automation, and data analysis — so if your task isn't programming, the same setup still applies.

Connecting models and tools

Before goose is useful, you'll typically want to connect two things: a model and whatever tools or data it should reach.

Choosing an LLM

goose works with 15+ providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and Bedrock. You can authenticate either with API keys or by using existing Claude, ChatGPT, or Gemini subscriptions via ACP (Agent Client Protocol). If you want to run models locally, Ollama is among the supported providers.

Adding extensions

Extensions connect goose to external systems — databases, APIs, browsers, GitHub, Google Drive, and more — through the Model Context Protocol (MCP), an open standard. The project documents 70+ extensions and lets you add skills or build your own extension.

A few capabilities worth knowing about as you set up:

  • MCP Apps — extensions can render interactive UIs (buttons, forms, visualizations) directly inside goose Desktop.
  • Subagents — you can spawn independent subagents to handle tasks in parallel, such as code review, research, or file processing, keeping the main conversation clean.
  • Recipes — workflows can be captured as portable YAML configs that include instructions, extensions, parameters, and subrecipes, then shared with a team or run in CI.

Editor and agent interoperability

goose acts as an ACP server, so you can connect to it from Zed, JetBrains, or VS Code. It can also use ACP agents like Claude Code and Codex as providers — meaning goose can sit on either side of that protocol.

Where to get help

The project points to three community channels, each suited to a different need:

Channel What it's for
Discord Asking questions, sharing what you've built, and getting help from the community
GitHub Starring, forking, filing issues, and contributing code — goose is built in the open
Extensions directory Browsing community-built MCP extensions and adding your own

If you hit a problem during setup, GitHub issues are the place for reproducible bugs, while Discord is better for "how do I do X" questions and sharing working configurations.

Governance and long-term viability

goose is part of the Agentic AI Foundation at the Linux Foundation. The project states this ensures it remains vendor-neutral, community-governed, and open for the long term — relevant if you're evaluating whether to build on it rather than just try it.

A practical first session

A reasonable first pass looks like this:

  1. Open the site's Get started section and pick desktop app, CLI, or API based on your workflow.
  2. Install and launch goose on your platform.
  3. Connect one LLM provider — an API key or an existing subscription via ACP.
  4. Add one extension relevant to your first task (for example, a browser or GitHub extension).
  5. Try a small, concrete task end to end before layering on Recipes or subagents.
  6. If something breaks, check GitHub issues; if you need guidance, ask in Discord.

The site does not publish pricing details in the material reviewed here, so treat cost as something to confirm per provider rather than assuming goose itself is free or paid.