Goose

Goose is an open-source, extensible AI agent introduced by Block (the company behind Square and Cash App) in January 2025. It goes beyond code suggestions — Goose can install dependencies, execute code, edit files, and run tests autonomously using any LLM you choose. It runs entirely on your machine, making it a privacy-first alternative to cloud-based agents.

With roughly 35,000 GitHub stars, 1,700+ MCP extensions, and support for 25+ LLM providers, Goose has grown into one of the most popular open-source AI agent frameworks — now governed by the Linux Foundation.

Key Features

1. General-Purpose Agent

Not just for code — Goose handles research, writing, automation, and data analysis. It reads and writes files, runs code and tests, installs dependencies, and iterates on output until tasks are done.

2. Native Desktop, CLI, and API

• Desktop app: macOS, Linux, and Windows.
• Full CLI: For terminal workflows.
• API: Embed goose anywhere in your tools and products.
• Built in Rust for performance and portability.

3. Model Context Protocol (MCP) Native

Block co-developed the MCP protocol with Anthropic. Goose connects to 1,700+ MCP extensions — GitHub, Figma, Docker, and more — and can discover new systems on the fly.

4. Any LLM You Want

Works with 25+ LLM providers, including Claude, GPT, and local models. Run entirely offline with a local model for complete data privacy — your code never leaves your hardware.

5. Extensible Architecture

Modular design with custom distributions: preconfigure providers, extensions, and branding to build your own goose distro.

Use Cases

• Software Engineering: Reduce time on maintenance and repetitive tasks; search, navigate, and write code autonomously.
• Automation: Script and automate workflows with the computer controller.
• Data Analysis: Analyze data and generate reports with local models.
• Privacy-Sensitive Work: Full local operation with open-source LLMs.

Pricing

Goose is free and open source (Apache 2.0). You pay only for the LLM API usage of the provider you connect — or use a local model at zero marginal cost.

Advantages & Unique Selling Points

1. Privacy by Design: Everything runs on your machine; data stays under your control.
2. MCP Pioneer: Direct lineage to the protocol's development, with the largest extension ecosystem.
3. Model Freedom: Choose any LLM — frontier models or fully local ones.
4. Cross-Platform: Desktop, CLI, and embeddable API in one Rust codebase.

User Reviews

"Goose is an on-machine AI agent and it's open source — everything is stored on your computer, private, you have control over the data." — Rizel Scarlett, Block Open Source Team

"Open source using Goose with an open-source LLM is so powerful — you can own your data and you don't even need internet." — The Square Developer Podcast

"35,000 stars, 1,700 MCP extensions, built in Rust, now governed by the Linux Foundation." — Prism Labs, YouTube

Getting Started

1. Install: Download the desktop app for macOS/Linux/Windows, or install the CLI: curl -fsSL .../download_cli.sh | bash.
2. Configure: On first launch, pick your LLM provider and enter credentials.
3. Enable Extensions: Toggle built-in extensions or add MCP servers from the community.
4. Start Automating: Give goose a task and let it plan, execute, and iterate.

Limitations

• Local Hardware Demands: Local LLMs need sufficient RAM/GPU; a 64GB machine works, but more helps.
• Extension Quality Varies: Not all community MCP extensions are equally reliable.
• Configuration Learning Curve: Setting up providers and extensions takes some effort.
• Autonomy Trade-Offs: Autonomous execution requires careful permission management.

Conclusion

Goose is a compelling, privacy-first open-source AI agent from a major fintech company. Its MCP-native architecture, 25+ provider support, and desktop/CLI/API surface make it both powerful and flexible. If you want an AI agent that runs locally, respects your data, and can be extended to almost any tool, Goose is an excellent choice — especially for privacy-conscious teams.

Recommended for: Developers and teams wanting a local-first, open-source AI agent with MCP extension support.

Not recommended for: Users who want a fully managed cloud agent with no setup or infrastructure concerns.