AI

Goose: The Open-Source AI Agent Reshaping Agentic Development with 44.7k+ GitHub Stars

Goose is a production-ready, open-source AI agent supporting 15+ LLM providers and 70+ MCP extensions. Desktop app, CLI, and API for code, automation, and research.

Tosin Akinosho

— 4 min read

Goose is a general-purpose, open-source AI agent that runs natively on your machine—not just for code, but for research, writing, automation, data analysis, and any task you need to accomplish. With 44.7k+ GitHub stars and active development from the Agentic AI Foundation (AAIF) at the Linux Foundation, Goose represents a mature, production-ready alternative to proprietary coding agents. It works with 15+ LLM providers and connects to 70+ extensions via the Model Context Protocol (MCP), making it the most extensible AI agent framework available today.

What is Goose?

Goose is a native desktop application (macOS, Linux, Windows), a full-featured CLI, and an embeddable API—all built in Rust for performance and portability. Originally developed as an internal tool at Block (the company behind Square and Cash App), Goose was open-sourced in 2025 and subsequently donated to the Agentic AI Foundation, ensuring long-term community governance and development.

Unlike single-purpose coding assistants, Goose is a general-purpose agent that can handle complex workflows across multiple domains. It integrates with any LLM provider—Anthropic Claude, OpenAI GPT, Google Gemini, Ollama, OpenRouter, Azure, AWS Bedrock, and more—giving you flexibility to choose your preferred model or use your existing subscriptions via the Anthropic Cloud Platform (ACP).

The project is actively maintained with commits within hours of this writing, 474 contributors, and 132 releases. It's the reference implementation for the Model Context Protocol (MCP), meaning Goose shapes the future of how AI agents connect to external tools and data sources.

Core Features and Architecture

Multi-Provider LLM Support — Goose works with 15+ LLM providers out of the box. Switch between Claude, GPT-4, Gemini, or local models (Ollama) without changing your workflow. Use API keys directly or authenticate via ACP for seamless integration with your existing subscriptions.

Model Context Protocol (MCP) Integration — Connect to 70+ extensions via MCP, the open standard for AI agent tool integration. MCP servers expose capabilities like GitHub access, Slack integration, database queries, file operations, and custom business logic. Goose is the reference implementation, meaning new MCP features are tested and validated in Goose first.

Native Desktop Application — A full-featured UI for macOS, Linux, and Windows. Manage sessions, view agent reasoning, inspect tool calls, and control execution—all from a native app. The desktop experience is polished and production-ready, not a web wrapper.

Powerful CLI — For terminal-first developers, Goose includes a comprehensive CLI that supports all desktop features. Run agents in CI/CD pipelines, automate workflows, and integrate Goose into your existing tooling.

Extensible Architecture — Built in Rust with TypeScript for the UI, Goose is designed for extensibility. Create custom MCP servers, build skill recipes, and distribute your own Goose distros with preconfigured providers and branding.

Session Management — Goose maintains persistent sessions, allowing agents to learn from previous interactions and maintain context across multiple runs. Sessions can be saved, loaded, and shared for reproducibility.

Recipe System — Define reusable workflows as recipes. Goose includes built-in recipes for code review, release risk assessment, and common development tasks. Create custom recipes for your team's specific workflows.

Get free AI agent insights weekly

Join our community of builders exploring the latest in AI agents, frameworks, and automation tools.

Join Free

Getting Started

Installation is straightforward. Download the desktop app from goose-docs.ai for your platform, or install the CLI:

curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash

Prerequisites:

  • An LLM API key (Claude, OpenAI, etc.) or ACP authentication
  • macOS 11+, Linux (Ubuntu 20.04+), or Windows 10+
  • For local inference: Ollama or compatible runtime

First Run: Launch the desktop app or run goose in your terminal. Configure your LLM provider, and you're ready to start. The quickstart guide at goose-docs.ai/docs/quickstart walks you through your first agent interaction in under 5 minutes.

Example: Code Review Agent

goose run --recipe code-review --file src/main.rs

This command runs Goose's built-in code review recipe on your file, analyzing it for bugs, performance issues, and best practices.

Real-World Use Cases

Autonomous Code Review and Refactoring — Use Goose to review pull requests, suggest refactorings, and identify security issues before they reach production. The code review recipe integrates with GitHub via MCP, allowing Goose to fetch PRs, analyze diffs, and post comments automatically.

Data Analysis and Research — Goose can process large datasets, generate reports, and conduct research across multiple sources. Connect it to your data warehouse via MCP, and let it explore, analyze, and summarize findings—all without manual intervention.

CI/CD Pipeline Automation — Embed Goose in your CI/CD workflows to automate testing, deployment validation, and release risk assessment. The release risk check recipe evaluates changes for potential issues before deployment.

Documentation Generation — Goose can read your codebase, understand its architecture, and generate comprehensive documentation. Use it to keep docs in sync with code changes automatically.

How It Compares

vs. Claude Code (Anthropic) — Claude Code is a terminal-based agent optimized for coding tasks with superior codebase understanding. Goose is more general-purpose and works with any LLM provider, giving you flexibility. Claude Code has tighter integration with Claude's capabilities; Goose prioritizes extensibility via MCP.

vs. Cursor — Cursor is an IDE with AI features built-in. Goose is a standalone agent that can be embedded anywhere. Cursor excels at interactive coding; Goose excels at autonomous workflows and automation. They serve different use cases—Cursor for interactive development, Goose for automation and general tasks.

vs. AutoGen (Microsoft) — AutoGen is a Python framework for building multi-agent systems. Goose is a complete application with UI, CLI, and API. AutoGen requires more setup and coding; Goose works out of the box. Both are powerful, but Goose is more accessible for non-developers.

Strengths: Open-source, multi-provider support, MCP integration, native apps, active development, Linux Foundation backing.

Limitations: Newer than some competitors (though mature), smaller ecosystem than proprietary tools, requires some technical setup for advanced customization.

What's Next

The Goose roadmap includes enhanced vision/image support for local inference models, cross-platform improvements, and deeper integrations with enterprise tools. The project is actively exploring advanced agentic capabilities like multi-step reasoning, improved error recovery, and better handling of long-running tasks.

As part of the Agentic AI Foundation, Goose will continue to evolve as the reference implementation for MCP, ensuring that new standards and capabilities are tested and validated in production. The community is growing rapidly, with contributions from developers worldwide building custom MCP servers and Goose distros for specialized use cases.

Goose represents a turning point in open-source AI development: a mature, production-ready agent that doesn't lock you into a single provider or vendor. Whether you're automating code reviews, conducting research, or building complex workflows, Goose gives you the flexibility and power to do it your way.

Sources

Read more