# Goose — Open source AI coding agent with a desktop app, CLI, and 70+ MCP extensions across 15+ LLM providers, backed by the Linux Foundation

> Source: The Agents Index — https://theagentsindex.com/goose (structured, researched, re-verified)
> Facts last verified: 2026-08-20

Goose is an open source AI coding agent distributed as a native desktop app for macOS, Linux, and Windows, plus a command line interface. It automates software development tasks by connecting to over 70 Model Context Protocol extensions and more than 15 LLM providers.

Goose is stewarded by the Agentic AI Foundation (AAIF), a project under the Linux Foundation, and is aimed at developers who want an extensible, provider-agnostic coding agent rather than a closed SaaS tool.

| Fact | Value |
| --- | --- |
| Website | https://goose-docs.ai |
| Pricing | Free and open source (Apache 2.0) |
| API | Yes |
| Best for | Devs who want an open source coding agent that reuses an existing Claude or ChatGPT plan |
| Not for | Non-technical users or teams that want a fully managed, zero-configuration coding assistant with vendor support included; Goose requires choosing, configuring, and paying for your own LLM provider separately. |

## Pricing

| Tier | Price |
| --- | --- |
| Open source | Free (self-hosted) |

## Verdict

Goose is a genuinely open, provider-agnostic alternative to closed coding agents: install it free under Apache 2.0, point it at any of 15+ LLM providers, and extend it with 70+ MCP tools rather than being stuck with one vendor's model and roadmap. Backing from the Linux Foundation's Agentic AI Foundation and a community of 500+ contributors gives it more institutional weight than a typical side project.

The tradeoff is that Goose is a toolkit, not a managed product. There's no bundled subscription or support line, users configure their own provider and absorb that provider's costs and rate limits directly, and evaluation currently relies almost entirely on the vendor's own documentation. For developers comfortable managing their own API keys and provider choices, that's a fair exchange for the flexibility it buys.

## How it works

1. Install Goose as a native desktop app (macOS, Linux, Windows) or via the CLI
2. Connect an LLM provider such as Anthropic, OpenAI, or OpenRouter using an API key
3. Start a session and give Goose a coding task; it plans, executes, and calls MCP extensions as needed
4. Resume a prior session at any time with goose session -r, or run Goose as an ACP server inside Zed, JetBrains, or VS Code

## Who it's for

- Developers who want a free, open source coding agent instead of a paid closed-source SaaS tool
- Engineering teams that want to embed agent behavior into internal tools via API rather than use a standalone app
- IDE users on Zed, JetBrains, or VS Code who want agent capability inside their existing editor via ACP

## Strengths and weaknesses

- ✓ Free and open source under Apache 2.0, with no single-vendor lock-in on the tool itself
- ✓ Works with 15+ LLM providers, letting users pick models on cost or capability grounds
- ✓ Backed by the Linux Foundation's Agentic AI Foundation with a large contributor base (45k+ stars, 500+ contributors)
- ✓ Extensible via 70+ MCP extensions and usable inside popular IDEs through ACP server mode
- ✗ Being self-hosted, users must bring and pay for their own LLM API keys; there is no bundled subscription covering model usage
- ✗ Getting started requires choosing and configuring an LLM provider, which adds setup friction compared with all-in-one hosted coding assistants
- ✗ No named enterprise support tier or SLA is documented, which may matter to teams used to vendor-backed SaaS agents
- ⚠ Details here come from the vendor's own documentation site (goose-docs.ai) with one supporting GitHub issue in the project's own repository; independent third-party reviews of Goose are not available yet.
- ⚠ No published pricing tiers exist to compare, since the product itself is free and open source; only underlying LLM provider costs vary.

## Key features

- **15+ LLM providers** — Works with Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, Bedrock, ChatGPT Codex, Claude Code CLI, and Tetrate Agent Router Service, so teams are not locked to one model vendor.
- **70+ MCP extensions** — Connects to the coding agent's environment through the Model Context Protocol, including a built-in Computer Controller extension for webscraping, file caching, and automation.
- **Desktop app and CLI** — Native desktop app on macOS, Linux, and Windows, plus a full CLI, so the same agent can run interactively or scripted.
- **ACP server mode** — Runs as an Agent Client Protocol server, letting it be driven from inside Zed, JetBrains IDEs, or VS Code instead of a separate app.
- **Embeddable API** — Provides an API for embedding Goose's agent behavior into other applications rather than only using the desktop or CLI front ends.
- **Session security controls** — Includes prompt injection detection, tool permission controls, and a sandbox mode to constrain what an agent session can execute.

## Use cases

- **Terminal-first coding automation** — A developer installs the Goose CLI and starts a resumable session to automate repetitive coding tasks like refactors, test writing, or debugging without leaving the terminal.
- **Agent inside an existing IDE** — A JetBrains, Zed, or VS Code user connects Goose as an ACP server so agent assistance appears inside the editor they already use, instead of switching to a separate desktop app.
- **Embedding an agent into internal tooling** — An engineering team uses Goose's API to embed its agent behavior into an internal developer platform or CI pipeline rather than exposing the standalone app to every engineer.
- **Provider cost and performance tradeoffs** — A team routes Goose through OpenRouter for access to 200+ pay-per-use models, or tries Tetrate Agent Router Service to use the $10 in free credits granted on first authentication.

## Integrations

Anthropic API · OpenAI · Google · Ollama · OpenRouter · Azure OpenAI · Amazon Bedrock · ChatGPT Codex · Claude Code CLI · GitHub Copilot · Tetrate Agent Router Service · Zed · JetBrains IDEs · VS Code · Computer Controller extension

## Sources

- https://goose-docs.ai
- https://goose-docs.ai/docs/quickstart
- https://goose-docs.ai/docs
- https://github.com/aaif-goose/goose/issues/887