Goose is an open-source AI agent framework developed by Block that enables developers to connect large language models to real-world actions including building projects from scratch, writing and executing code, debugging failures, and orchestrating complex workflows. It goes beyond code suggestions by acting as a true autonomous agent that reads and writes files, runs code and tests, installs dependencies, and handles additional actions as needed, all in real time within the development environment. Goose was designed to be the most interoperable and extensible AI coding agent available, built on the open Model Context Protocol developed in collaboration with Anthropic.
Goose works with any LLM and supports multi-model configuration to optimize performance and cost, with seamless MCP server integration for connecting to any external system through simple extensions. It is available as both a desktop application and a CLI, supports named sessions with full chat history, subagents for parallel task execution, and custom skills for adding project-specific context. Unlike cloud-based AI solutions, Goose can operate entirely locally on a developer's machine, keeping sensitive code and data private. Its modular design means developers can extend its capabilities without modifying the core framework.
Goose is designed for developers who want a flexible, open-source AI agent that can be customized to fit any workflow and integrated with any toolchain through the Model Context Protocol. It is particularly valuable for teams that use multiple AI models and want the ability to switch between or combine them based on task requirements and cost considerations. Compared to proprietary AI coding agents, Goose offers complete transparency and control over the agent's behavior, making it the natural choice for developers and organizations that prioritize open-source tooling and local-first development practices.