Goose

Goose is a native, open-source AI agent designed to actually get work done rather than simply generate responses. It runs locally on your computer and is available as a desktop application, command-line interface and API.

Originally created by Block, Goose is now part of the Agentic AI Foundation under the Linux Foundation and is developed as an open-source project under the Apache 2.0 license.

Unlike traditional AI coding assistants that primarily provide suggestions, Goose can perform multi-step tasks. It can work with files, execute commands, modify code, run tests, perform research, analyze data and automate workflows.

One of Goose's biggest strengths is its model-agnostic architecture. Instead of forcing users into a single AI provider, Goose supports 15+ providers, including OpenAI, Anthropic, Google, Ollama, OpenRouter, Azure and Amazon Bedrock. Users can also use existing Claude, ChatGPT or Gemini subscriptions through ACP.

Goose is also deeply integrated with the Model Context Protocol (MCP). Its extension ecosystem provides connections to services such as databases, APIs, browsers, GitHub and Google Drive, allowing the agent to interact with external systems and execute real-world workflows. The official site currently documents 70+ MCP extensions.

Users can additionally create reusable workflows using Recipes, use subagents for parallel tasks, and interact with MCP Apps that can render interactive interfaces inside the Goose desktop application.

For users who care about control and flexibility, Goose is particularly interesting because it runs on the user's machine, supports multiple model providers, and is not tied to a proprietary AI ecosystem.