Installation & First Run

Goose is a Rust-based binary distributed through shell installers and package managers. It runs on Linux, macOS, and Windows. This lesson covers installation, model configuration, and completing your first task with Goose.

What you'll learn

• Install via shell script: curl -fsSL https://goose.ai/install | bash
• Goose works with any LLM provider through API keys or OAuth
• First run walks through model selection and provider configuration

Build it

Step 1: Install

Run the official installation script to download and unpack the latest release.

curl -fsSL https://goose.ai/install | bash

The installer downloads the binary, places it in ~/.local/bin/, and adds it to your PATH.

Step 2: Configure a provider

Launch the interactive configuration wizard to select your LLM provider and supply the necessary credentials.

goose configure

Select your provider (OpenAI, Anthropic, Google, or OpenRouter) and enter your API key. Goose stores credentials locally.

Step 3: Run your first task

Trigger a single-shot execution to test that the model can access your local filesystem and run commands.

goose run "Check the disk usage on this machine and summarize what's consuming the most space"

Goose executes the task using its built-in terminal and file tools, or connects to relevant MCP servers if configured.

Step 4: Interactive mode

Start the persistent REPL loop if you prefer a continuous conversation over single-shot commands.

goose

Starts a REPL session. Use /help to see available commands.

What goes wrong

Mistake	How you notice it	The fix
API key not configured	Goose fails with authentication error	Run goose configure and verify the key is set correctly
Binary not found	Command not found after install	Source your shell config or add ~/.local/bin to PATH

Next: MCP Extensions