block/goose

Goose is an open-source AI agent built by Block (the company behind Square and Cash App) that doesn't just suggest code — it executes entire workflows autonomously. Install packages, write and run code, debug failures, manage files, orchestrate multi-step tasks across APIs. All locally on your machine, with any LLM you choose. What separates Goose from the growing pile of AI coding assistants is extensibility through MCP (Model Context Protocol). Over 3,000 tools and APIs connect through MCP servers, turning Goose into whatever kind of agent you need. Connect it to your database, your CI/CD pipeline, your Slack workspace, your cloud infrastructure. The agent adapts to your toolchain instead of forcing you into theirs. Written in Rust (58.3% of the codebase), Goose is fast and resource-efficient. It runs as both a desktop app and a CLI, supports multi-model configurations for balancing performance against cost, and ships under Apache 2.0 — so you can deploy it commercially without licensing headaches. The growth trajectory tells the story: from launch in January 2026 to 37,000+ GitHub stars in under 3 months, with 4,079 commits and hundreds of contributors. Block's Open Source Program Office backs the project with dedicated engineering resources. Goose was featured on the Sequoia Capital podcast and donated alongside Anthropic's MCP to the Agentic AI Foundation (AAIF). The "recipes" system lets you save and replay multi-step workflows. Set up a recipe for "initialize a new microservice with tests, CI config, and deployment manifest" and run it across every new project. The agent handles the execution while you review the results. Key limitation: Goose is a power-user tool. The learning curve is steeper than Cursor or GitHub Copilot. You need to understand MCP server configuration, model selection trade-offs, and agentic workflow patterns to get full value. If you want a simple autocomplete, look elsewhere. If you want an agent that can run your entire development pipeline, Goose is the open-source answer. Related: Cursor AI Editor | Open Source AI Agents in 2026