Command Palette

Search for a command to run...

GitHubGitHub

Goose Installation Guide

Install Block's open-source Goose agent as a desktop app or CLI, connect it to Anthropic, OpenAI, Google or a local model, and extend it with 70+ MCP extensions.

Goose is a general-purpose AI agent that runs on your machine — not just for code, but for research, writing, automation and data analysis. It ships as a desktop app, a CLI and an API, and it is part of the Agentic AI Foundation at the Linux Foundation.

System Requirements

  • macOS, Linux or Windows on a 64-bit processor
  • 4 GB RAM or more
  • An API key for one of the supported providers, or a local model through Ollama
  • git for the code-related workflows

Installation Instructions

The Desktop App

Download the app for macOS, Linux or Windows from goose-docs.ai and install it like any other desktop application.

The CLI

# macOS, Linux and WSL
curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash
 
# Verify
goose --version
# Build from source
git clone https://github.com/aaif-goose/goose.git
cd goose
make build
make install

Configuration

Goose needs a provider before it can do anything:

# Start the configuration flow
goose configure
 
# Or set the key directly
export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
export GOOGLE_API_KEY=...
ProviderEnvironment variable
AnthropicANTHROPIC_API_KEY
OpenAIOPENAI_API_KEY
GoogleGOOGLE_API_KEY
OpenRouterOPENROUTER_API_KEY
AzureAZURE_OPENAI_API_KEY
Amazon BedrockAWS_BEARER_TOKEN_BEDROCK
Ollama (local)OLLAMA_HOST

Extensions

Extensions are the tools Goose can call. They are shared through MCP:

# List the available extensions
goose extension list
 
# Add one
goose extension add developer
 
# Remove one
goose extension remove developer
# Configure a custom MCP server
goose mcp add filesystem -- /usr/bin/npx -y @modelcontextprotocol/server-filesystem .

Using Goose

# Start an interactive session
goose session
 
# Run a single task non-interactively
goose run "summarise the changelog for the last release"
 
# List the available recipes
goose recipe list
> use the developer extension
> add the "fix flaky test" recipe

Troubleshooting

Best Practices

  1. Install only the extensions you need — each one adds tools the model can call.
  2. Use a local model for sensitive code — Ollama keeps everything on the machine.
  3. Check the debug log first when something misbehaves; GOOSE_DEBUG=1 is detailed.
  4. Use recipes for repeatable multi-step tasks.
  5. Keep the API key in a secret store, not in a shell profile on a shared machine.
  6. Pin the version in CI images so an update cannot break a pipeline.
  7. Build a custom distribution if you need preconfigured providers and branding.

Next Steps

Aider

Install Aider with the one-line installer, uv or pipx, connect it to a model provider and let it commit your changes to git as it works.

OpenCode

Install the open-source OpenCode terminal agent with the install script, npm or Homebrew, connect a model provider and bootstrap the project with an AGENTS.md file.