
Goose
Open-source agent extensible via MCP.
Goose is an open-source AI agent by Block that runs locally and is extensible via the Model Context Protocol (MCP).
Transparency Note: This page may contain affiliate links. We may earn a commission at no extra cost to you. Learn more.
Overview
Goose: The Open Source Agent by Block (2026 Review)
Rating: 9.0/10 (Best Open Source CLI Agent)
1. Executive Summary
Goose is an open-source AI agent developed by Block (Square). It is designed to be an extensible, developer-focused agent that runs in your terminal or on your desktop. Unlike closed-source agents, Goose is built to be hacked on and extended via the Model Context Protocol (MCP).
2. Core Features
2.1 Model Context Protocol (MCP) Support
- Extensibility: Goose can be extended with "tools" that allow it to interact with any API, database, or local application.
- Community Tools: A growing ecosystem of MCP servers allows Goose to control Slack, Jira, GitHub, and more.
2.2 Local & Secure
- BYOK (Bring Your Own Key): You use your own API keys for LLMs.
- Local Execution: Goose runs locally, giving it full access to your filesystem and local tools (with your permission).
3. Pros & Cons
Pros
- Fully Open Source: MIT licensed.
- MCP Ecosystem: Unlimited potential for integration.
- Backed by Block: Strong engineering support.
Cons
- UI: Primarily CLI/Desktop focused, less visual than some IDE agents.
- Setup: Requires more technical know-how than a "one-click" SaaS tool.
5. Conclusion
Goose is the hacker's agent. If you want to build your own AI workflows and integrations, Goose provides the perfect foundation.
Use Cases
Custom workflows
Local automation
Tool integration



