Goose AI

Goose is an open-source, extensible AI coding agent developed by Block (the company behind Square and Cash App) that goes far beyond simple code suggestions. Unlike traditional AI coding assistants that operate inside an editor as autocomplete tools, Goose is a fully autonomous agent capable of installing dependencies, executing shell commands, editing files, running tests, debugging errors, and orchestrating multi-step engineering workflows on a developer's local machine. It is designed to handle the messy, end-to-end reality of software engineering rather than just generating code snippets in isolation.

One of Goose's defining strengths is its model-agnostic architecture. Developers can plug in virtually any large language model — Anthropic Claude, OpenAI GPT models, Google Gemini, local models served through Ollama, or models from Groq, Databricks, OpenRouter, and others. This freedom lets engineers optimize for cost, latency, privacy, or capability depending on the task. The tool runs entirely on the user's own infrastructure, which means source code never has to leave the local environment unless the user explicitly chooses a cloud-hosted model provider.

Goose is built around the Model Context Protocol (MCP), the open standard introduced by Anthropic for connecting AI agents to external tools and data sources. Through MCP, Goose can integrate with file systems, GitHub, Jira, browsers, databases, design tools, and any other service that exposes an MCP server. This extensibility makes Goose effectively unbounded: developers and teams can build or install custom MCP servers — called "extensions" in Goose terminology — to give the agent new abilities specific to their stack.

The project ships in two forms: a command-line interface for terminal-first developers and a desktop application for those who prefer a graphical chat UI. Both share the same underlying agent engine and configuration. Sessions are persistent, so a developer can resume an in-progress refactor, debugging session, or exploration after closing the app. Goose also supports recipes — reusable, shareable workflow templates — that allow teams to codify repeatable engineering tasks such as onboarding scripts, code reviews, or release checklists.

Because it is genuinely open source under the Apache 2.0 license and free to use, Goose has become an attractive alternative for engineers and organizations who want the productivity of agentic AI tooling without vendor lock-in, per-seat subscriptions, or mandatory cloud data transfer. It is particularly popular among developers who already work with multiple LLM providers, security-conscious teams handling sensitive codebases, and tinkerers who want to extend their tooling rather than be confined to a closed ecosystem.