Goose
The free, open-source AI agent you run yourself, as a desktop app or CLI
Editorial take
Goose gives you agentic capabilities close to Claude Code for free: the agent costs nothing and you pay only for the model you connect, whether that is an API key, a local model or an existing Claude, ChatGPT or Gemini subscription. Originally built by Block and now governed by the Linux Foundation's Agentic AI Foundation under Apache 2.0, it is the strongest free alternative to paid coding agents. Note that autonomous mode is the default, so switch to manual or smart approval if you want to review actions first.
What is Goose?
Goose is an open-source AI agent that runs on your own machine. It was originally built by Block (the company behind Square and Cash App), released in early 2025, and contributed to the Linux Foundation's Agentic AI Foundation (AAIF) in December 2025, where it is now developed as a vendor-neutral, community-governed project. It is written in Rust and licensed under Apache 2.0.
Goose ships as a native desktop app for macOS, Linux and Windows, a full CLI, and an API for embedding it in other applications. It can read and edit files, run shell commands, and carry out multi-step tasks such as scaffolding projects, refactoring, running tests and fixing bugs. The project also positions it for non-coding work like research, writing, automation and data analysis. Extensions use the Model Context Protocol (MCP), with 70+ documented extensions, and repeatable workflows can be saved as YAML recipes and shared.
Goose works with dozens of model providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure and Amazon Bedrock. Instead of API keys, you can use an existing Claude, ChatGPT or Gemini subscription through the Agent Client Protocol (ACP), and the desktop app can run local models in-process without Ollama. The software itself is free; you pay only for the model provider you choose. Approval modes range from fully autonomous (the default) to manual approval and smart approval, which runs low-risk actions automatically and asks before higher-impact ones.
Best for
Developers and technical users who want a free, vendor-neutral agent for coding and other tasks, using models or subscriptions they already have
Key strength
Full agentic capabilities for free, with any model provider
Score breakdown (out of 5)
What you would use it for
- →Running multi-file coding tasks autonomously without a paid agent subscription
- →Choosing any AI model (Claude, GPT, Gemini or a local model) as the backend for agentic tasks
- →Keeping code on your machine by running the agent entirely locally
- →Automating repetitive development tasks like schema migrations or scaffolding
- →Setting up and running tests as part of a complete agentic development session
Pros & Cons
👍 Pros
- ✓Free and open source - you only pay your model provider
- ✓Can use an existing Claude, ChatGPT or Gemini subscription instead of per-token API billing
- ✓Wide model choice, including local models
- ✓Desktop app means you do not need to work in the terminal
- ✓Vendor-neutral governance under the Linux Foundation's Agentic AI Foundation
- ✓Large MCP extension ecosystem
👎 Cons
- ✗You have to choose and configure a model provider yourself
- ✗API costs accumulate on large tasks
- ✗Autonomous mode is the default, so it edits files and runs commands without asking unless you switch modes
- ✗Built-in local models are slow to load and have small context windows
Key Features
- ✓ Native desktop app for macOS, Linux and Windows, plus a CLI and an embeddable API
- ✓ Reads and edits files and runs shell commands
- ✓ Dozens of model providers, including Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure and Amazon Bedrock
- ✓ Uses existing Claude, ChatGPT or Gemini subscriptions through ACP
- ✓ Built-in local model inference in the desktop app
- ✓ Autonomous, manual approval, smart approval and chat-only modes
- ✓ 70+ documented MCP extensions, including MCP Apps with interactive UIs
- ✓ Reusable YAML recipes and parallel subagents
- ✓ Works as an ACP agent in Zed, JetBrains IDEs and VS Code
- ✓ Prompt injection detection, sandbox mode and an adversary reviewer
- ✓ Open source (Apache 2.0 license)
Available on
Integrates with
Goose Pricing
✅ Goose has a free plan - no credit card required to start.
Open Source
- ✓Desktop app, CLI and embeddable API
- ✓Bring your own API keys or use an existing Claude, ChatGPT or Gemini subscription
- ✓Dozens of model providers plus local models
- ✓70+ documented MCP extensions
- ✓Recipes and parallel subagents
- ✓Apache 2.0 license
Goose vs Competitors
AI models powering Goose
From the blog

GitHub Actions configs from Anthropic, Google, and OpenAI all vulnerable to same RCE
Security researchers discovered that default GitHub Actions configurations published by Anthropic, Google, and OpenAI for their coding agents could be exploited via unauthenticated GitHub issues to achieve remote code execution.
Sep 14, 2026

OpenAI agents attacked RubyGems without disclosure. Supply chain security concerns mount.
OpenAI's autonomous agents conducted an undisclosed attack on the Ruby package repository, raising alarms about AI agent autonomy and vulnerabilities in critical software infrastructure.
Sep 12, 2026

AI Workflows Promise Automation. Reality Demands Expertise.
A Reddit discussion surfaces a persistent pain point: multi-agent AI systems look seamless in demos but demand significant setup complexity in practice. Users question whether the operational burden outweighs actual productivity gains.
Sep 6, 2026
Developer resources
Related Tools
Claude is the best AI assistant for long-document work and research where accuracy matters. Its context window of up to 1M tokens and resistance to sycophancy make it the honest choice when you need a thinking partner rather than an agreeable assistant. Start on the free tier; the Pro plan at $20/month is worth it if you regularly hit the free usage limits, and Max ($100/month) or Ultra ($200/month) suit heavier daily use.
ChatGPT is the most capable all-in-one AI assistant available. Its combination of image generation, voice mode, code execution, web browsing, and a large custom GPT ecosystem makes it the default choice for users who want one tool that handles everything. The tradeoff is that it is rarely the deepest tool on any single task.
Cursor is the most powerful AI code editor available. Its multi-file editing is ahead of plugin-based alternatives, making large-scale refactors practical in ways GitHub Copilot cannot match. At $30/month for Premium (Grok models included, other models billed at API rates) it is a meaningful upgrade for any developer doing more than small isolated changes.
GitHub Copilot is the right starting point for developers adopting AI-assisted coding. It integrates into whatever editor you already use, requires no workflow change, and has a usable free tier. The Pro plan at $10/month is the lowest-cost path to unlimited AI completions, though chat and agent work now draw from a monthly credit allowance; heavy agent users should look at Pro+ ($39) or Max ($100).
This page contains affiliate links. We may earn a commission at no extra cost to you. Learn more.