Block · Tools

Goose

Visit website

At a glance

Price
Pricing on request
Vendor
Block

Specifications & properties

Key decision factors

Pricing model
Open source 1
Tool form
  • CLI agent
1
Autonomy level
Autonomous agent 1
Model choice
Bring your own key (BYO) 1

Pricing

Price from
Pricing on request 1
Free tier
Yes 1

Capabilities

Terminal/command execution
Yes 1
Report data / suggest a correction

Goose is Block's open-source, extensible AI agent that runs on your machine and can install, execute, edit, and test with any LLM.

Profile

Goose is a free, open-source AI agent that runs on your own machine and handles coding, research, writing, automation, and data-analysis tasks, originally built by Block and now governed by the Agentic AI Foundation (AAIF) at the Linux Foundation.

Who builds it

Goose started as a project inside Block (the fintech company behind Square and Cash App) and has since moved under the Linux Foundation's Agentic AI Foundation, with more than 500 contributors maintaining it as a community-driven, vendor-neutral project rather than a single company's closed product.

Core features

  • Runs locally: available as a native desktop app, a command-line tool, and an API, so agent execution happens on your machine rather than solely in a vendor's cloud.
  • Bring-your-own model: works with 15+ LLM providers including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and AWS Bedrock, so you can reuse an existing Claude, ChatGPT, or Gemini subscription instead of paying twice.
  • MCP extensions: connects to 70+ Model Context Protocol extensions covering databases, APIs, browsers, GitHub, and Google Drive, and can render interactive UIs from those extensions via "MCP Apps."
  • Recipes: shareable, YAML-based workflow definitions that let a team package a repeatable agent task and hand it to someone else.
  • Subagents: Goose can spin off independent subagents to work on parts of a task in parallel.
  • Security features: prompt-injection detection, granular tool permissions, a sandbox mode, and an "adversary reviewer" designed to catch risky agent actions before they run.

Pricing

Goose itself is fully open source and free to use, licensed under Apache 2.0. There is no Goose subscription or seat fee — the only ongoing cost is whatever you pay the LLM provider(s) you connect it to (an API key, or an existing paid Claude/ChatGPT/Gemini plan), which is entirely separate from Goose's own pricing.

Who it's for

Developers and teams who want a vendor-neutral, locally-run agent framework instead of a closed SaaS coding assistant — especially those who already have an LLM provider relationship and want to plug it into a flexible, MCP-extensible agent rather than being locked into one vendor's chat UI. Its recipe and subagent system also suits teams that want to standardize and share repeatable automation workflows, while the sandboxing and permission controls make it usable in security-conscious environments.

Frequently asked questions

Is Goose free?

Yes, Goose is fully open source and free to use under the Apache 2.0 license. The only cost you incur is for whichever LLM provider you connect it to — your own API key or an existing paid subscription like Claude, ChatGPT, or Gemini — Goose itself has no subscription fee.

Who builds Goose?

Goose was originally created by Block, the fintech company behind Square and Cash App, and is now governed by the Agentic AI Foundation (AAIF) under the Linux Foundation, with more than 500 community contributors maintaining it.

Which LLM providers does Goose support?

Goose works with 15+ providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, and AWS Bedrock, and it's designed so you can reuse an existing Claude, ChatGPT, or Gemini subscription rather than paying separately.

Does Goose run locally or in the cloud?

Goose runs on your own machine as a native desktop app, a command-line tool, or via an API, on macOS, Linux, or Windows. Execution happens locally rather than exclusively inside a vendor's cloud, though the LLM call itself goes to whichever provider you've configured.

What is MCP support in Goose, and what can extensions do?

Goose connects to more than 70 Model Context Protocol (MCP) extensions covering databases, APIs, browsers, GitHub, and Google Drive, letting an agent pull in outside tools and data. Some extensions can also render interactive UIs directly in the Goose desktop app through a feature called MCP Apps.

Does Goose have any built-in safety or security controls?

Yes. Goose includes prompt-injection detection, granular tool permissions, a sandbox mode, and an adversary reviewer intended to catch risky agent actions before they execute, which is relevant for teams that want to run an autonomous agent without giving it unchecked access.