# Goose — AI Developer Agent by Block > Open-source AI developer agent by Block (Square). Goose automates coding tasks with extensible toolkits, session memory, and MCP server support in your terminal. ## Install Save as a script file and run: ## Quick Use ```bash brew install block/goose/goose # or pipx install goose-ai goose session start ``` Start chatting — Goose reads your codebase, runs commands, and edits files. ## What is Goose? Goose is an open-source AI developer agent created by Block (the company behind Square and Cash App). It runs in your terminal, understands your project context, and can write code, run tests, manage git, and interact with external services through extensible toolkits and MCP servers. **Answer-Ready**: Goose is an open-source AI developer agent by Block that automates coding tasks in your terminal. It features extensible toolkits, session memory, MCP server support, and works with Claude, GPT, and local models. 10k+ GitHub stars. **Best for**: Developers who want an open-source terminal AI agent with enterprise backing. **Works with**: Claude, GPT-4o, Gemini, Ollama. **Setup time**: Under 2 minutes. ## Core Features ### 1. Extensible Toolkits Goose ships with built-in toolkits and supports custom ones: ``` Built-in toolkits: - developer: File operations, code editing, git - screen: Screenshot and UI interaction - repo-context: Codebase understanding - jira: Issue tracking integration - github: PR and issue management ``` ### 2. MCP Server Support ```yaml # ~/.config/goose/profiles.yaml default: provider: anthropic model: claude-sonnet-4-20250514 mcpServers: memory: command: npx args: ["-y", "@modelcontextprotocol/server-memory"] github: command: npx args: ["-y", "@modelcontextprotocol/server-github"] ``` ### 3. Session Memory Goose remembers context within and across sessions: ```bash goose session start # New session goose session resume # Resume last session goose session list # List all sessions ``` ### 4. Profile System Configure different profiles for different projects: ```yaml # ~/.config/goose/profiles.yaml frontend: provider: anthropic model: claude-sonnet-4-20250514 toolkits: [developer, screen] backend: provider: openai model: gpt-4o toolkits: [developer, repo-context, jira] ``` ### 5. Multi-Provider Support | Provider | Models | |----------|--------| | Anthropic | Claude Sonnet, Opus | | OpenAI | GPT-4o, o1 | | Google | Gemini 2.5 | | Ollama | Llama, Mistral, local models | ## Example Session ``` $ goose session start Goose> Fix the failing tests in src/auth/ Reading test output... 3 failures found Analyzing src/auth/middleware.ts Fixing token validation logic Running tests... All 47 tests passing Committing: "fix: correct JWT token validation in auth middleware" ``` ## FAQ **Q: How does Goose compare to Claude Code?** A: Both are terminal AI agents. Goose is open-source with a toolkit plugin system. Claude Code has deeper Claude integration and broader tool access. They serve similar use cases. **Q: Is it production-ready?** A: Yes, built and used internally by Block (a Fortune 500 company). Active open-source development. **Q: Can I extend it with custom tools?** A: Yes, via custom toolkits (Python plugins) or MCP servers. ## Source & Thanks > Created by [Block](https://github.com/block). Licensed under Apache 2.0. > > [block/goose](https://github.com/block/goose) — 10k+ stars ## 快速使用 ```bash brew install block/goose/goose goose session start ``` 终端启动 Goose,开始对话式编码。 ## 什么是 Goose? Goose 是 Block(Square 母公司)开源的 AI 开发者代理。在终端运行,理解项目上下文,支持可扩展工具包和 MCP 服务器。 **一句话总结**:Goose 是 Block 开源的终端 AI 开发代理,支持可扩展工具包、会话记忆和 MCP 服务器,10k+ GitHub stars。 **适合人群**:需要有企业背书的开源终端 AI 代理的开发者。**支持**:Claude、GPT-4o、Ollama。 ## 核心功能 ### 1. 可扩展工具包 内置开发、截图、Git、Jira、GitHub 工具包,支持自定义。 ### 2. MCP 服务器支持 配置文件中声明 MCP 服务器,自动加载。 ### 3. 会话记忆 跨会话保持上下文,支持恢复历史会话。 ### 4. 多配置文件 不同项目使用不同模型和工具组合。 ## 常见问题 **Q: 和 Claude Code 比较?** A: 都是终端 AI 代理。Goose 开源有插件系统,Claude Code 有更深的 Claude 集成。 **Q: 生产就绪?** A: 是,Block(财富 500 强)内部使用并开源。 ## 来源与致谢 > [block/goose](https://github.com/block/goose) — 10k+ stars, Apache 2.0 --- Source: https://tokrepo.com/en/workflows/dedbb70b-7ebd-4987-9ba3-a883f45e8b5a Author: Agent Toolkit