Back to landing
IDE MCP install

Add Torus MCP to your IDE.

Pick your IDE below — Cursor, VS Code, Claude Desktop, Windsurf, or any MCP-compatible editor. After install, ask: “Use Torus to plan my idea.”

Choose your IDE

Install in Cursor

Config file: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)

Setup steps
  1. Click "Add to Cursor" or paste the config into .cursor/mcp.json
  2. Restart Cursor completely
  3. Open Settings → MCP and confirm torus-ai-builder is connected
  4. Ask: "Use Torus to plan my idea: [your app idea]"
Example prompt: “Use Torus to plan my idea: a Next.js SaaS dashboard with AI recommendations.”
mcp.json
http://localhost:3000/api/mcp
{
  "mcpServers": {
    "torus-ai-builder": {
      "url": "http://localhost:3000/api/mcp"
    }
  }
}

Idea input

Type your app, product, or MCP idea directly inside the IDE.

Features

Receive must-have and optional features with complexity and reasoning.

Architecture

Get stack, services, data, AI, auth, and integration guidance.

Development guide

Follow phased prompts built for AI coding tools.

Deployment guide

See hosting, env vars, checks, and production rollout steps.

Tool links

Use recommended AI tools and services for each project layer.

Deploy MCP

How to deploy the MCP server

Torus MCP ships with the app — no separate MCP hosting. Deploy the Next.js app and the /api/mcp endpoint goes live automatically.

Option A — Remote HTTPRecommended

Best for production and team sharing

  1. Deploy Torus to Vercel: npx vercel --prod
  2. Add env vars in Vercel (Firebase, Groq, SearXNG, NEXT_PUBLIC_APP_URL)
  3. Verify: curl https://your-app.vercel.app/api/mcp
  4. Add the remote mcp.json config to Cursor or VS Code
  5. Ask your IDE: “Use Torus to plan my idea”
Option B — Local stdio

Best for local dev when Torus runs on localhost

  1. Start Torus: npm run dev
  2. Use the stdio config below in .cursor/mcp.json
  3. Or run manually: npm run mcp
  4. Set TORUS_BASE_URL to your Vercel URL to point at production
Remote HTTP mcp.json (production)
{
  "mcpServers": {
    "torus-ai-builder": {
      "url": "http://localhost:3000/api/mcp"
    }
  }
}
Local stdio mcp.json (development)
{
  "mcpServers": {
    "torus-ai-builder": {
      "command": "node",
      "args": [
        "mcp/torus-mcp-server.mjs"
      ],
      "env": {
        "TORUS_BASE_URL": "http://localhost:3000"
      }
    }
  }
}

Full guide: mcp/README.md in the repo.