Goose (Block)

What is Goose (Block)?

Goose (Block) is a local AI agent for developers that handles code, research, writing, automation, and data analysis from one workspace. It spans a desktop app, CLI, and API, with 70+ MCP extensions, Recipes, MCP Apps, and subagents for parallel work. Goose also connects to 15+ model providers and ACP environments like Zed, JetBrains, and VS Code.

Last verifiedMay 17, 2026How we evaluate

Visit Goose (Block)

At a glance

Best for: Goose is best for developers who want a local AI agent that fits terminal, desktop, and API workflows.

What does Goose (Block) do?

Goose runs locally as a general-purpose AI agent that can take on code, research, writing, automation, and data analysis from one workspace. Its desktop app, CLI, and API let you move between interactive use, terminal workflows, and embedded automation without changing tools. The extensible layer connects to 70+ documented MCP extensions, while Recipes package repeatable workflows into portable YAML that can be shared, run in CI, and extended with instructions and parameters. At scale, goose is backed by 38k+ GitHub stars and 400+ contributors, with support for 15+ model providers and ACP connections to coding environments like Zed, JetBrains, and VS Code. It also supports subagents for parallel task handling and MCP Apps for interactive UIs inside the desktop experience. The project is part of the Agentic AI Foundation, and the docs show local execution, open standards, and a broad ecosystem rather than a single-model or single-editor workflow.

Why use Goose (Block)?

Open standards support means you can connect through MCP and ACP instead of locking into one editor or model stack.
Recipes turn agent work into portable YAML, making repeatable processes easier to share and automate in CI.
Subagents and MCP Apps let teams split work into parallel tasks and build richer interactive agent tools.
Broad provider support lets you use API keys or existing Claude, ChatGPT, or Gemini subscriptions.

Who is Goose (Block) for?

Developers who want an agent that works across desktop, terminal, and embedded workflows.
Engineering teams who need repeatable AI workflows they can share and run in CI.
Tool builders who want to extend an agent with MCP apps, extensions, and subagents.
Teams using multiple model providers who want to keep existing subscriptions and keys.

What are Goose (Block)'s key features?

Desktop app, CLI, and API

Use Goose from a desktop app, command line, or API to run code and workflows across GitHub and Google Drive without changing tools.

Extensible

Extend Goose with 70+ MCP extensions and custom workflows, so teams can adapt the agent to internal systems and repeatable tasks.

Any LLM, including your subscriptions

Connect 15+ providers such as Anthropic, OpenAI, Azure, Bedrock, Ollama, OpenRouter, Claude, ChatGPT, and Gemini to fit existing model access.

Recipes

Reuse prebuilt recipes for common tasks, giving teams a faster starting point for automation instead of building every workflow from scratch.

MCP Apps

Run MCP apps inside Goose to connect external tools and data sources, including GitHub and Google Drive, in a standard agent workflow.

Subagents

Split work into subagents so Goose can handle multi-step jobs in parallel and keep larger coding or workflow tasks organized.

Security

Use Goose with security controls designed for agentic work, helping teams manage access while connecting to providers like Azure and Bedrock.

What does Goose (Block) integrate with?

GitHub
Google Drive
Anthropic
OpenAI
Google
Ollama
OpenRouter
Azure
Bedrock
Claude
ChatGPT
Gemini
Zed
JetBrains
VS Code
Claude Code
Codex

What are Goose (Block)'s use cases?

Developer workflow across surfaces

Developers who want an agent that works across desktop, terminal, and embedded workflows use Goose to move from quick local prompts to repeatable coding tasks without changing tools. They rely on Desktop app, CLI, and API for the same agent experience while keeping their preferred model setup with Any LLM, including your subscriptions.

Repeatable CI recipes

Engineering teams who need repeatable AI workflows they can share and run in CI use Goose to package common tasks into Recipes. They can standardize review, triage, or code-generation flows, then run them consistently through the CLI or API instead of rebuilding prompts each time.

Extend agents with MCP apps

Tool builders who want to extend an agent with MCP apps, extensions, and subagents use Goose to connect specialized capabilities into one workflow. With MCP Apps and Subagents, they can assemble custom agent behavior for internal tools, then keep those workflows organized through the Extensible architecture.

Multi-provider model control

Teams using multiple model providers who want to keep existing subscriptions and keys use Goose to route work across providers without locking into one stack. Any LLM, including your subscriptions lets them choose Anthropic, OpenAI, Gemini, or local models like Ollama while maintaining Security controls around how the agent operates.

How does Goose (Block) work?

Install Goose in the Desktop app, CLI, or API, then connect your first workflow where you already work. Start with a local task so you can see the agent handle code or operations end to end.
Choose your model provider in Any LLM, including your subscriptions, and point Goose at the keys you already use. Switch between Anthropic, OpenAI, Gemini, Ollama, or OpenRouter as needed.
Turn a recurring task into a Recipe so the steps are reusable. Save the prompt, inputs, and expected output, then run it again from the CLI or share it with your team.
Add MCP Apps and Subagents when a workflow needs specialized tools or narrower responsibilities. Extend Goose with the integrations you need, then chain agents together for more complex jobs.
Use Security settings to control what the agent can access and how it behaves. Keep refining your workflows as your team adopts them, so the same setup stays reliable in daily use.

Frequently asked questions

What is Goose (Block)?

Goose (Block) is a local AI agent for developers that handles code, research, writing, automation, and data analysis from one workspace. It spans a desktop app, CLI, and API, with 70+ MCP extensions, Recipes, and subagents for parallel work. Goose also connects to 15+ model providers and ACP environments like Zed, JetBrains, and VS Code.

What is Goose (Block) used for? Who is it for?

Goose (Block) is used for Desktop app, CLI, and API, Extensible, and Any LLM, including your subscriptions. It's built for Developers, Engineering teams, and Tool builders.