# 安装 Marrow AI: Control Layer for AI Agents

- 类型: MCP 工具
- 说明: Approval gate and audit trail for AI agent actions: policy check first, proof after. Any MCP client.
- 仓库: https://github.com/getmarrow/marrow-mcp

## 方式一 · 一键代装（复制提示词让 AI 代劳）

### Claude Code
```
请帮我安装 askskill 上的 "Marrow AI: Control Layer for AI Agents" MCP 服务：
执行：claude mcp add 'ai-getmarrow-marrow' -- npx -y @getmarrow/mcp
```

### Codex
```
请帮我安装 askskill 上的 "Marrow AI: Control Layer for AI Agents" MCP 服务：
在 ~/.codex/config.toml 中添加：
[mcp_servers.ai-getmarrow-marrow]
command = "npx"
args = ["-y","@getmarrow/mcp"]
然后重启 Codex 生效
```

### Cursor
```
请帮我安装 askskill 上的 "Marrow AI: Control Layer for AI Agents" MCP 服务：
在 .cursor/mcp.json 的 "mcpServers" 中添加：
"ai-getmarrow-marrow": { "command": "npx", "args": ["-y","@getmarrow/mcp"] }
然后在 Cursor 设置 > MCP 中启用
```

## 方式二 · 命令行安装（原生命令与配置）

### Claude Code
```bash
claude mcp add --env MARROW_API_KEY=YOUR_VALUE --transport stdio ai-getmarrow-marrow -- npx -y @getmarrow/mcp
```
需要环境变量：MARROW_API_KEY

### Codex
```bash
codex mcp add --env MARROW_API_KEY=YOUR_VALUE ai-getmarrow-marrow -- npx -y @getmarrow/mcp
```
配置文件 ~/.codex/config.toml：
```toml
[mcp_servers.ai-getmarrow-marrow]
command = "npx"
args = ["-y","@getmarrow/mcp"]

[mcp_servers.ai-getmarrow-marrow.env]
MARROW_API_KEY = "YOUR_VALUE"
```
需要环境变量：MARROW_API_KEY

### Cursor
配置文件 .cursor/mcp.json：
```json
{
  "mcpServers": {
    "ai-getmarrow-marrow": {
      "command": "npx",
      "args": [
        "-y",
        "@getmarrow/mcp"
      ],
      "env": {
        "MARROW_API_KEY": "YOUR_VALUE"
      }
    }
  }
}
```
一键安装：cursor://anysphere.cursor-deeplink/mcp/install?name=ai-getmarrow-marrow&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsIkBnZXRtYXJyb3cvbWNwIl0sImVudiI6eyJNQVJST1dfQVBJX0tFWSI6IllPVVJfVkFMVUUifX0
需要环境变量：MARROW_API_KEY