MCP Servers

What is MCP

MCP (Model Context Protocol) enables the AI to connect with external services and tools. Through MCP servers, the AI can search the web, automate browsers, call APIs, and more.

Built-in MCP Servers

SoAgents includes 6 pre-configured MCPs:

Name	Function	Cost
Context7	Auto-fetch latest library/framework docs	Free
Playwright	Browser automation, screenshots, form filling	Free
DuckDuckGo	Web search	Free
Tavily Search	AI-optimized search, structured results	1000 free/month
Gemini Image	Image generation and editing	Free tier
Edge TTS	Text-to-speech, 400+ voices	Free

Enabling MCP

1. Open Settings → MCP
2. Find the MCP you want to enable
3. Toggle the switch
4. Some MCPs require an API Key (e.g., Tavily)

Once enabled, the AI automatically uses the corresponding tools in conversations.

Adding Custom MCP

If built-in MCPs aren’t enough, add your own:

1. Settings → MCP → Add Custom
2. Choose type:
   • STDIO — Local CLI tools (e.g., npx, uvx)
   • HTTP — Remote HTTP service
   • SSE — Server-Sent Events service
3. Configure command/URL and arguments
4. Set environment variables if needed

Example: Adding a STDIO MCP

{
  "command": "npx",
  "args": ["-y", "@anthropic/mcp-server-xxx"],
  "env": {
    "API_KEY": "your-key"
  }
}

Discover More MCPs

Find more MCP servers at:

• mcp.so — MCP server directory
• smithery.ai — MCP community

Troubleshooting

”Command not found” after enabling MCP

Make sure the required runtime is installed (e.g., Node.js, Python) and available in your system PATH.

MCP connection timeout

Check your network connection, or configure a proxy for HTTP-type MCPs.