# MCP Server 集成

桥接服务器可被 AI Agent 通过 [MCP (Model Context Protocol)](https://modelcontextprotocol.io) 调用，实现「AI 写作 → 一键分发到多平台」。

## 桥接 HTTP API

桥接在 `http://<host>:9601` 暴露简单 HTTP 接口，任何语言都能调用：

| 方法 | 路径 | 说明 |
|------|------|------|
| GET  | `/status`    | 桥接与扩展连接状态 `{connected, mode, uptime}` |
| GET  | `/platforms` | 各平台登录状态（含缓存）`{platforms[], cached, cachedAt}` |
| POST | `/request`   | 转发任意请求到 Chrome 扩展 |

### POST /request

```json
{
  "method": "sync",           // 或 listPlatforms / auth 等扩展支持的方法
  "params": { "platforms": "zhihu,csdn" },
  "token": "<Token>"
}
```

响应：

```json
{ "result": { ... } }
```

## MCP 工具封装示例

用 Node 写一个最小 MCP Server，把桥接封装成工具：

```js
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

const BRIDGE = process.env.BRIDGE_URL || "http://host.docker.internal:9601";

const server = new Server({ name: "wechatsync-publisher", version: "1.0.0" },
  { capabilities: { tools: {} } });

server.setRequestHandler("tools/list", async () => ({
  tools: [{
    name: "list_platforms",
    description: "列出已登录的内容平台",
    inputSchema: { type: "object", properties: {} }
  }, {
    name: "publish_article",
    description: "把 Markdown 文章同步到指定平台",
    inputSchema: {
      type: "object",
      properties: {
        file: { type: "string", description: "本地 Markdown 文件路径" },
        platforms: { type: "string", description: "逗号分隔平台 id" }
      },
      required: ["file", "platforms"]
    }
  }]
}));

server.setRequestHandler("tools/call", async (req) => {
  if (req.params.name === "list_platforms") {
    const r = await fetch(`${BRIDGE}/platforms`);
    const d = await r.json();
    return { content: [{ type: "text", text: JSON.stringify(d.platforms, null, 2) }] };
  }
  if (req.params.name === "publish_article") {
    const { file, platforms } = req.params.arguments;
    const r = await fetch(`${BRIDGE}/request`, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({ method: "sync", params: { file, platforms } })
    });
    const d = await r.json();
    return { content: [{ type: "text", text: JSON.stringify(d) }] };
  }
});

await server.connect(new StdioServerTransport());
```

## Claude Desktop / 其他客户端配置

```json
{
  "mcpServers": {
    "wechatsync": {
      "command": "node",
      "args": ["/path/to/wechatsync-mcp/index.js"],
      "env": { "BRIDGE_URL": "http://host.docker.internal:9601" }
    }
  }
}
```

## 典型工作流

1. Agent 调用 `list_platforms` → 获取已登录平台
2. Agent 生成/改写文章，写入 Markdown
3. Agent 调用 `publish_article` → 桥接转发给 Chrome 扩展 → 扩展在各平台创建草稿
4. 返回草稿编辑链接给用户

无 Chrome 扩展连接时，桥接返回缓存的平台状态，Agent 可据此判断是否需要提示用户登录。