MCP Protocol Integration: Connecting Custom External Tools & APIs
How Unchained Code supports Model Context Protocol (MCP) servers via /api/mcp-check and /api/mcp-update-config.
Modern AI developers need their coding agents to query live PostgreSQL databases, fetch Figma design tokens, execute specialized CLI utilities, and interact with private corporate APIs. Unchained Code implements native support for the open Model Context Protocol (MCP), enabling seamless plug-and-play tool extensions.
Unchained Code lets you configure custom MCP servers in Settings > MCP Config. The builder validates endpoints via /api/mcp-check, discovers available tool schemas dynamically, and injects tool calling definitions directly into DeepSeek-V4 Pro and Qwen 3.8 Max runtime sessions.
Step-by-Step Walkthrough & User Guide
Access MCP Configuration Panel
Navigate to Settings > Model Context Protocol (MCP).
Add Remote or Stdio MCP Server
Enter your MCP server URL or SSE endpoint (e.g., http://localhost:3001/sse).
Automated Schema Discovery & Validation
Unchained Code pings /api/mcp-check and retrieves tool declarations automatically.
Invoke Tools Naturally via AI Prompts
Prompt the agent: 'Query our local Postgres database for user stats'. The agent calls the MCP tool seamlessly.
💡Pro Tips & Engineering Best Practices
- •Use MCP servers to connect live Supabase or Neon database schemas directly into chat context.
- •Local stdio MCP stubs run safely without exposing private endpoints over the public web.
- •Save and export your MCP configurations as JSON across your development team.
Code & Technical Integration
{
"mcpServers": {
"postgres": {
"url": "http://localhost:3001/sse",
"capabilities": ["resources", "tools"]
}
}
}