Detailed overview

Goose is an open-source, local-first AI agent built in Rust, available as a native desktop app (macOS, Linux, Windows), a CLI, and an API. Originally created by Block, it is now stewarded by the Agentic AI Foundation under the Linux Foundation. Goose connects large language models to files, tools, and external services through the Model Context Protocol (MCP), supporting 70+ extensions such as databases, GitHub, browsers, and Google Drive. It works with 15+ model providers, including the ability to use existing Claude, ChatGPT, or Gemini subscriptions via the Agent Client Protocol (ACP). Developers can capture workflows as portable YAML 'recipes,' spawn parallel subagents for tasks like code review or research, and build interactive MCP-based UIs. Security features include prompt-injection detection, tool permission controls, and sandboxing. While extensible and inspectable, Goose requires technical setup and is not limited to coding—also supporting writing, automation, and data analysis workflows.

Pricing details

Free and open source (Apache 2.0); users supply their own model provider API keys or can use existing Claude, ChatGPT, or Gemini subscriptions via ACP.

Visit Goose

Key features

Local agent
Tool extensions
Workflow automation
Model flexibility

Systems you can build

CLI tools
Backend services
APIs
Automation scripts
Internal developer tools
Data pipelines

Strengths

Open source
Composable tools
Good local control

Limitations

Technical setup
Provider costs vary
Smaller community

Verification sources

Listing details are checked against official sources.

https://block.github.io/goose/