MCP Setup Guide

Inject WeCom development knowledge into your AI via MCP (Model Context Protocol) — the most elegant integration method.

Why MCP?

Compared to manually copying SKILL files, MCP offers three key advantages:

• Zero clone — No repository download, no file copying
• On-demand loading — AI only reads the SKILL it needs, avoiding context overload
• Auto-update — Package updates deliver the latest knowledge automatically

---

Universal Configuration

All MCP-compatible AI tools share the same core JSON config:

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

No API Key, no Docker, no server required.

Just install Node.js 22+ and you're good to go.

---

Tool-Specific Setup

Claude Desktop

Claude Desktop is Anthropic's desktop client.

Config file location:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

Save and restart Claude Desktop.

---

Claude Code

Claude Code is Anthropic's CLI coding assistant.

One-line setup:

claude mcp add omnisockit -- npx @omnisockit/mcp-server

Verify:

claude mcp list
# You should see omnisockit registered

---

Cursor

Cursor is an AI-native code editor.

Config file location: .cursor/mcp.json in project root (project-level) or global settings.

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Windsurf

Windsurf is Codeium's AI IDE.

Config file location: ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

VS Code + GitHub Copilot

MCP support in VS Code via GitHub Copilot.

Config file location: .vscode/mcp.json in project root.

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Trae

Trae is an AI IDE by ByteDance.

Config file location: .trae/mcp.json in project root.

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Antigravity

Antigravity is Google DeepMind's desktop AI coding assistant.

Via Settings: Settings → MCP Servers → Add the following config:

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Cline

Cline is a VS Code extension.

Config file location: Cline MCP settings (cline_mcp_settings.json)

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Roo Code

Roo Code is a VS Code AI extension.

Config file location: Roo Code MCP settings (mcp_settings.json)

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Kilo Code

Kilo Code is a VS Code AI extension.

Config file location: Kilo Code MCP settings (mcp_settings.json)

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Continue

Continue supports VS Code and JetBrains IDEs.

Config file location: ~/.continue/config.yaml

models:
  - name: omnisockit
    provider: mcp
    mcpServers:
      - name: omnisockit
        command: npx
        args:
          - "@omnisockit/mcp-server"

---

Cherry Studio

Cherry Studio is a desktop AI client.

GUI Setup:

1. Open Cherry Studio → Settings → MCP Servers
2. Click "Add"
3. Fill in:
   • Name: omnisockit
   • Command: npx
   • Args: @omnisockit/mcp-server
4. Save

---

Zed

Zed is a high-performance code editor.

Config file location: ~/.config/zed/settings.json

Zed uses context_servers (not mcpServers):

{
  "context_servers": {
    "omnisockit": {
      "command": {
        "path": "npx",
        "args": ["@omnisockit/mcp-server"]
      }
    }
  }
}

---

OpenAI Codex CLI

OpenAI Codex CLI is OpenAI's command-line coding tool.

Option 1: CLI argument

codex --mcp-server "npx @omnisockit/mcp-server"

Option 2: Config file

{
  "mcpServers": {
    "omnisockit": {
      "command": "npx",
      "args": ["@omnisockit/mcp-server"]
    }
  }
}

---

Verification

After setup, type in your AI chat:

Help me implement WeCom OAuth login

If configured correctly, the AI will automatically read the wecom-auth SKILL, and its response will include:

• Precise OAuth URL format
• The 5-minute code expiration limit
• redirect_uri encoding requirements

---

How It Works

┌──────────────────────────────────────────┐
│          Your AI Tool                    │
│  (Claude / Cursor / VS Code / ...)      │
│                                         │
│  You: "Help me implement WeCom OAuth"   │
│                                         │
│  AI determines it needs wecom-auth      │
│  ↓ Reads via MCP Protocol               │
└──────────────┬──────────────────────────┘
               │ stdio (JSON-RPC)
┌──────────────▼──────────────────────────┐
│       OmniSocKit MCP Server             │
│                                         │
│  41 SKILLs → 41 MCP Resources          │
│  AI reads on-demand, not all at once    │
│                                         │
│  Returns wecom-auth SKILL content       │
└─────────────────────────────────────────┘
               ↓
  AI generates precise code based on SKILL

Key features:

• 🔒 Zero risk — No external API calls, knowledge-only
• ⚡ Zero config — No API Key, no server needed
• 🎯 On-demand — AI reads only the SKILLs it needs

---

Security

Architecture Safety

OmniSocKit MCP Server uses a knowledge-only architecture, fundamentally different from most MCP Servers:

                  ╔═══════════════════════════════════╗
                  ║  Typical MCP Server (calls APIs)   ║
                  ║                                    ║
                  ║  AI → MCP → External API → Data    ║
                  ║  ⚠️ Needs creds · Network · Risk   ║
                  ╚═══════════════════════════════════╝

                  ╔═══════════════════════════════════╗
                  ║  OmniSocKit MCP Server (read-only) ║
                  ║                                    ║
                  ║  AI → MCP → Local SKILL → Text     ║
                  ║  ✅ No creds · No network · Safe    ║
                  ╚═══════════════════════════════════╝

Five Security Guarantees

Guarantee	Details
🚫 No external API calls	Never sends HTTP requests to WeCom, WeChat, or any third-party server. Zero outbound network traffic after process startup
🔑 No credentials required	Does not ask for corpid, secret, access_token, or any sensitive information. No environment variables needed
📊 No data collection	No telemetry, no analytics, no usage tracking. Reports nothing to any server
🏠 Runs locally only	Communicates with AI tools via stdio (standard I/O). Listens on no TCP/UDP ports, starts no HTTP server
📖 Verifiable content	All SKILL knowledge is based on the Open WeCom Skills open-source knowledge base. Sources are fully transparent

How to Verify?

If you'd like to verify the security claims yourself:

# 1. Download the npm package locally (no install)
npm pack @omnisockit/mcp-server

# 2. Extract and inspect the source
tar -xzf omnisockit-mcp-server-*.tgz
ls package/

# 3. Search for network calls (should find none)
grep -r "fetch\|axios\|http\|request\|net\." package/dist/

TL;DR

OmniSocKit MCP Server = a process that reads local files. It reads compiled SKILL documents and returns them to AI via stdio. No network, no credentials, no side effects.

---

FAQ

npx command not found?

Make sure Node.js 22+ is installed. npx is included with Node.js.

Slow first startup?

npx downloads the @omnisockit/mcp-server package on first run. Subsequent starts will be faster.

My AI tool is not listed?

Any tool supporting the MCP protocol can be integrated. The config format is usually similar to the examples above.

---

More Platforms

OmniSocKit MCP Server currently provides all 41 WeCom SKILLs for free. More social platform development knowledge is coming soon.