Skip to content

ask-experts-mcp

MCP server that routes domain-specific questions to expert LLMs (Hunyuan, Qwen, etc.) via OpenAI-compatible APIs

build.protocol_tooling.mcp#mcp#model-context-protocol#claude#llm#hunyuan#qwen#doubao

Install ask-experts-mcp in your MCP client

ask-experts-mcp is a Model Context Protocol server. Add it to your MCP client config once, restart, and the server's tools become available to your AI assistant. The same JSON snippet below works across all four major clients — only the config file path differs.

  1. Locate your client's MCP config file.
    • Cursor: ~/.cursor/mcp.json
    • Claude Desktop (macOS): ~/Library/Application Support/Claude/claude_desktop_config.json
    • Claude Desktop (Windows): %APPDATA%\Claude\claude_desktop_config.json
    • VS Code: Settings → Extensions → MCP
    • Windsurf: Settings → MCP Servers
  2. Add ask-experts-mcp to the mcpServers map — paste the snippet below into your config file. If you already have other MCP servers, merge the entry into the existing mcpServers object.
  3. Restart your client so it picks up the new server.
  4. Verify — ask the assistant to list available tools; ask-experts-mcp's tools should appear.
{
  "mcpServers": {
    "ask-experts-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "ask-experts-mcp"
      ]
    }
  }
}

Install ask-experts-mcp in Cursor

Open ~/.cursor/mcp.json in your editor, paste the snippet above into mcpServers, save, and restart Cursor. ask-experts-mcp will show up in the assistant's tool list on next launch.

Install ask-experts-mcp in Claude Desktop (macOS)

Open ~/Library/Application Support/Claude/claude_desktop_config.json in your editor, paste the snippet above into mcpServers, save, and restart Claude Desktop (macOS). ask-experts-mcp will show up in the assistant's tool list on next launch.

Install ask-experts-mcp in Claude Desktop (Windows)

Open %APPDATA%\Claude\claude_desktop_config.json in your editor, paste the snippet above into mcpServers, save, and restart Claude Desktop (Windows). ask-experts-mcp will show up in the assistant's tool list on next launch.

Install ask-experts-mcp in VS Code

Open Settings → Extensions → MCP in your editor, paste the snippet above into mcpServers, save, and restart VS Code. ask-experts-mcp will show up in the assistant's tool list on next launch.

npm package: ask-experts-mcp

Transport

ask-experts-mcp supports the following MCP transports. Most AI clients use stdio by default for locally-installed servers.

stdio

From the ask-experts-mcp README

# ask-experts-mcp > 让任意支持 MCP 的客户端(Claude Code / Cursor / Cline / Continue / Zed 等)在遇到自己不擅长的领域知识时,**主动咨询训练数据更对口的国产专家模型**。 写微信小程序问 `wx.login`、写阿里云 OSS 直传、写飞书机器人——这些**中文互联网生态特有的领域**,海外训练为主的模型(Claude / GPT / Gemini)覆盖比国产模型薄。`ask-experts-mcp` 给调用方模型暴露一组按"擅长领域"命名的工具(`ask_hunyuan`、`ask_qwen`、`ask_doubao`),让它**自己读工具描述、自己挑该问谁**——不需要切 chat 窗口、复制粘贴。 | 领域 | 谁更熟 | |---|---| | 微信生态(小程序、公众号、企业微信、视频号、QQ、腾讯云) | 混元 | | 阿里云 / 钉钉 / 通义系产品 / 1688 / 淘宝 | Qwen | | 抖音 / 飞书 / 火山引擎 / 字节系开放平台 | 豆包 | 本工具只做两件事:**路由**(通过工具描述告诉调用方"这种问题该找谁")+ **接入**(任意 OpenAI 兼容端点,baseURL + model 配置化)。 不做的事:不做内置 LLM 自动选 expert(那会增加一次模型调用、和调用方判断打架);不做多模型对比 / 投票;不做对话历史(每次 `ask_*` 单轮无状态);不做流式;不持有用户的 API Key,不做 SaaS 托管。 --- ## 快速开始 要求 Node.js ≥ 18。 ### 路径 A:零配置(推荐) 只要在 MCP 客户端的 `env` 里塞 API Key,**不需要写任何 JSON 配置文件**。`ask-experts-mcp

Other Model Context Protocol servers in the same space as ask-experts-mcp. Each one adds different capabilities to your AI assistant — pick based on the data sources or workflows you need.

Browse the full MCP server directory or use Stork's one-line install to let your agent pick the right server automatically.