# Goose — Open AI Agent for Desktop, CLI, and API

> Goose delivers a native desktop app, CLI, and API for open agent workflows, with broad provider support and MCP extensions.

## Install

Copy the content below into your project:

## Quick Use

1. Install the CLI:
   ```bash
   curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash
   ```
2. Launch Goose and connect your provider or ACP-backed subscription.
3. Verify:
   - Complete one local task and confirm an MCP extension is discoverable before you expand scope.

## Intro

Goose delivers a native desktop app, CLI, and API for open agent workflows, with broad provider support and MCP extensions.

- **Best for:** developers who want one native agent runtime across desktop, CLI, API, and MCP extensions
- **Works with:** desktop app, shell workflows, API embedding, 15+ model providers, 70+ MCP extensions
- **Setup time:** 10-20 minutes

## Practical Notes

- Quant: the README claims support for 15+ providers plus 70+ MCP extensions via the open protocol.
- Quant: Goose now lives under the Agentic AI Foundation at the Linux Foundation, which matters for governance and project continuity.

## Why it matters

Goose stands out because it ships multiple interaction surfaces from one codebase instead of asking users to choose between desktop polish and CLI composability.

- A Rust implementation improves portability and fits the repo's positioning as a native runtime rather than a browser-only shell.
- Provider flexibility includes direct API keys and subscription-style ACP flows, which broadens adoption paths.
- The foundation move is operationally relevant if your team values community governance over vendor ownership.

## Rollout pattern

- Use the CLI first even if you plan to adopt the desktop app, because CLI validation is easier to script and document.
- Test one provider path and one MCP extension path separately before you blend them in the same workflow.
- Record whether your team needs the desktop app, the CLI, or the API, because each surface implies different support obligations.

## Watchouts

The repo is in a transition from `block/goose` to `aaif-goose/goose`, so older links and instructions may still exist in the ecosystem; pin current URLs in internal docs.

### FAQ

**Q: Is Goose only a CLI?**
A: No. The README describes a native desktop app, a full CLI, and an embeddable API.

**Q: Why does the foundation move matter?**
A: It changes project governance and signals a broader community stewardship model.

**Q: What should I validate first?**
A: Install the CLI, finish one task, and confirm at least one MCP extension works with your chosen provider.

## Source & Thanks

> Source: https://github.com/aaif-goose/goose
> License: Apache-2.0
> GitHub stars: 45,070 · forks: 4,616

---

<!-- ZH -->

## 快速使用

1. 安装 CLI：
   ```bash
   curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash
   ```
2. 启动 Goose，并连接你的模型提供商或 ACP 订阅。
3. 验证：
   - 完成一个本地任务，并确认至少一个 MCP 扩展可以被发现后再扩大范围。

## 简介

Goose 同时提供桌面应用、CLI 与 API，并支持大量模型提供商与 MCP 扩展，适合想在本地用一套可移植运行时承载代码、研究和自动化任务并统一操作入口、使用习惯和维护方式的开发者们。

- **适合谁：** 希望在桌面、CLI、API 与 MCP 扩展间复用同一套原生 Agent 运行时的开发者
- **可搭配：** 桌面应用、终端工作流、API 嵌入、15+ 模型提供商与 70+ MCP 扩展
- **准备时间：** 10-20 分钟

## 实战建议

- 量化信息：README 写明支持 15+ 模型提供商，并可连接 70+ MCP 扩展。
- 量化信息：项目已迁移到 Linux Foundation 下的 Agentic AI Foundation，这会影响治理和长期连续性。 

## 为什么值得收录

Goose 的亮点是它没有在“桌面体验”和“CLI 可组合性”之间二选一，而是把两者都放进同一套运行时里。

- Rust 实现让它更符合“原生运行时”的定位，而不是只靠浏览器壳子包装。
- 既支持 API key，也支持 ACP 这类订阅路径，模型接入方式更灵活。
- 迁移到基金会治理对重视长期中立性的团队来说是实际可感知的信号。

## 落地路径

- 即使最终准备上桌面版，也建议先验证 CLI，因为更容易写成脚本和运维文档。
- 模型 provider 路径和 MCP 扩展路径要分开验证，再合并到同一工作流里。
- 先明确团队真正需要桌面、CLI 还是 API，因为三种表面意味着不同支持成本。 

## 注意事项

项目从 `block/goose` 迁移到 `aaif-goose/goose` 的过程中，生态里还会残留旧链接；内部文档最好统一固定到新地址。

### FAQ

**Goose 只有 CLI 吗？**
答：不是。README 同时描述了原生桌面应用、完整 CLI 和可嵌入 API。

**为什么基金会迁移值得注意？**
答：它影响项目治理结构，也意味着更偏社区托管而不是单一公司控制。

**第一步该验证什么？**
答：安装 CLI，完成一个任务，并确认至少一个 MCP 扩展可用。

## 来源与感谢

> Source: https://github.com/aaif-goose/goose
> License: Apache-2.0
> GitHub stars: 45,070 · forks: 4,616


---
Source: https://tokrepo.com/en/workflows/goose-open-ai-agent-for-desktop-cli-and-api
Author: Script Depot