goose

goose is an Apache-2.0 open-source AI coding agent available as a desktop application and command-line tool for macOS, Linux and Windows. It works inside a developer's project environment, where it can inspect and change files, run commands and tests, and iterate on multi-step tasks rather than stopping at code suggestions. Users choose an LLM provider during setup, with documented options spanning hosted services, OpenAI-compatible endpoints and local model runtimes. The canonical project now lives in the aaif-goose/goose GitHub organization, and release downloads provide the supported desktop and CLI builds.

Top features

• Use the same agent from a graphical desktop interface or the terminal, with continuous sessions that retain the working conversation and project context.
• Connect to different model providers instead of locking the agent to one vendor. Provider configuration covers API-key services, compatible endpoints and local options, subject to each model's tool-use capabilities.
• Extend the agent through Model Context Protocol servers. Built-in and third-party extensions can add developer tools, browser control, databases, issue trackers and other external systems.
• Let the developer extension read and write files, execute shell commands, inspect project state and run tests. Because those tools can affect a machine or repository, goose exposes per-tool permission levels such as always allow, ask before and never allow.
• Package repeatable instructions, prompts, parameters and extension requirements as recipes. Recipes can be validated, shared by deep link and opened from the CLI or desktop app.
• Supply project-specific guidance through goose hints and use documented scheduling features for repeatable unattended workflows where the configured permissions and environment are appropriate.

Use cases

• Implement a feature across several files, run the relevant test suite and revise the change after failures.
• Investigate a codebase, explain unfamiliar components, prepare refactors or assist with dependency and configuration updates.
• Automate repository chores such as test generation, documentation updates, code review preparation and issue triage.
• Connect coding work to Git hosting, ticketing, databases or browser automation through MCP extensions.
• Turn a recurring engineering procedure into a parameterized recipe that teammates can review and reuse from either interface.