AI导哈

MCP服务器集合

一个集中管理自定义与第三方MCP服务器的集合，支持通过VS Code统一配置，用于将消息、邮件、知识库等外部服务接入AI编程助手。

  • by Mingz6
  • TypeScript
  • 本地部署
  • 更新于
在 GitHub 查看 Mingz6/mcp-servers

使用说明

mcp-servers

All MCP servers in one place — custom and third-party. VS Code's mcp.json points here for everything.

Server Registry

Server Type What it does Why I need it
teams-chat Custom (Node) Read/send Teams messages, extract PR links Standup, PR review requests, colleague replies
outlook Custom (Node) Read/send Outlook mail, search, mark read, attachments Email triage during standup, PO reply detection
imessage Custom (Python) Read iMessage/SMS via Messages.app, send via AppleScript Tenant/family chats
brain-mcp Custom (Node) Search/read/link/update ~/code/brain Markdown wiki Agent-readable knowledge layer for Copilot and Claude Code
discord-user Custom (Node) Read Discord servers/DMs via personal user token (read-only) See community channels the bot can't access
wechat Custom (Python) Query WeChat chats, search messages, voice transcription (read-only) Family/personal chat access from Copilot
flywheel Custom (Node) Track reference repos, check updates, scout for new patterns Daily self-improvement loop — knowledge aggregator for /flywheel skill
linkedin Custom (Python) Read/send LinkedIn messages, search people, view profiles Professional networking from Copilot
ms-loop Custom (Node) CRUD Microsoft Loop pages via Graph API Read/write Loop docs without switching apps
telegram Custom (Python) Read/send Telegram messages, search chats, list contacts Personal/group chat access via Telethon
whatsapp Custom (Node/JS) Read/send WhatsApp messages, search chats, list contacts Family/group chats via Baileys WebSocket
worker-center-mcp Custom (Node) Query worker-center status, price history, job pipeline Agent access to scheduled task dashboard
awesome-copilot Third-party (Docker) Microsoft MCP .NET sample — copilot instructions search Load custom copilot instructions dynamically
context7 Third-party (npx) Fetch up-to-date library docs from Upstash Get current API docs instead of outdated training data

Structure

mcp-servers/
├── packages/
│   ├── brain-mcp/         # Custom — local brain Markdown wiki API
│   ├── discord-user/      # Custom — Discord via personal user token
│   ├── flywheel/          # Custom — reference repo tracker
│   ├── imessage/          # Custom (Python) — iMessage/SMS via macOS
│   ├── linkedin/          # Custom (Python) — LinkedIn via browser automation
│   ├── ms-loop/           # Custom — Microsoft Loop via Graph API
│   ├── outlook/           # Custom — Outlook via Graph API
│   ├── teams-chat/        # Custom — MS Teams via Graph API
│   ├── telegram/          # Custom (Python) — Telegram via Telethon
│   ├── wechat/            # Custom (Python) — WeChat via SQLCipher
│   ├── whatsapp/          # Custom (Node/JS) — WhatsApp via Baileys
│   └── worker-center-mcp/ # Custom — worker-center dashboard API
├── scripts/
│   └── npx-nvm.sh         # Wrapper: loads nvm before running npx
└── README.md

Quick Start

npm install        # installs Node workspaces (all *.ts packages)
npm run build      # builds all Node servers

For Python servers (wechat, imessage, linkedin, telegram):

cd packages/wechat   # or imessage, linkedin, telegram
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

VS Code Config

All servers are registered in your VS Code mcp.json (user or workspace level). Each custom server uses a run.sh wrapper — no hardcoded Python/Node paths:

{
    "servers": {
        "teams-chat": {
            "type": "stdio",
            "command": "${userHome}/code/personal/mcp-servers/packages/teams-chat/run.sh",
            "env": {
                "TEAMS_MCP_CLIENT_ID": "your-client-id",
                "TEAMS_MCP_TENANT_ID": "your-tenant-id"
            }
        },
        "wechat": {
            "type": "stdio",
            "command": "${userHome}/code/personal/mcp-servers/packages/wechat/run.sh"
        },
        "brain": {
            "type": "stdio",
            "command": "${userHome}/code/personal/mcp-servers/packages/brain-mcp/run.sh",
            "env": {
                "BRAIN_MCP_ROOT": "${userHome}/code/brain"
            }
        }
    }
}

The wrappers handle runtime discovery (nvm for Node, .venv for Python) so upgrading Node/Python versions won't break your MCP config.

Adding a New Server

Custom server

  1. Create packages/{name}/ with its own package.json (Node) or requirements.txt (Python)
  2. Add a README.md in the package folder
  3. Add to the registry table above
  4. Add entry to VS Code's mcp.json

Third-party server

  1. Add entry to VS Code's mcp.json (use scripts/npx-nvm.sh for npx-based servers)
  2. Add to the registry table above with upgrade notes

Upgrading

Server How to upgrade
teams-chat cd packages/teams-chat && npm update && npm run build
outlook cd packages/outlook && npm update && npm run build
brain-mcp cd packages/brain-mcp && npm update && npm run build
wechat cd packages/wechat && .venv/bin/pip install --upgrade "mcp[cli]"
awesome-copilot Update the Docker image SHA in mcp.json
context7 Automatic — npx -y @upstash/context7-mcp always pulls latest

Troubleshooting

Server won't start in VS Code?

  • Check mcp.json paths are correct (use ${userHome} not ~)
  • Run the run.sh manually to test: ./packages/teams-chat/run.sh
  • For Python servers: check .venv exists — the wrapper will tell you how to create it
  • For npx servers: run scripts/npx-nvm.sh -y <package> manually to test

Auth issues?

  • Teams/Outlook: run the auth-test script in the package → npm run auth-test
  • WeChat: re-extract keys → sudo python3 packages/wechat/extract_key.py

相关 MCP