Vega-Lite MCP Server

# Deployment Guide (Future Enhancement) This guide is for deploying the Vega-Lite MCP Server as a **remote server** accessible over the internet. ⚠️ **Note**: The current implementation uses STDIO transport for local use. This guide describes the changes needed for remote deployment. ## Current Status: Local Only (STDIO) The server currently runs locally using: - **Transport**: `StdioServerTransport` (standard input/output) - **Clients**: Claude Desktop, VS Code (local) - **Deployment**: None needed - runs on user's machine ## To Deploy Remotely: Required Changes ### 1. Change Transport Layer You'll need to modify `src/index.ts` to support HTTP or SSE transport: ```typescript // Instead of StdioServerTransport import { SSEServerTransport } from "@modelcontextprotocol/sdk/server/sse.js"; // or import { StreamableHttpTransport } from "@modelcontextprotocol/sdk/server/http.js"; ``` ### 2. Add Web Server Create a new file `src/server.ts`: ```typescript import express from 'express'; import { Server } from '@modelcontextprotocol/sdk/server/index.js'; import { SSEServerTransport } from '@modelcontextprotocol/sdk/server/sse.js'; const app = express(); const PORT = process.env.PORT || 3000; app.get('/mcp', async (req, res) => { const transport = new SSEServerTransport('/mcp', res); await server.connect(transport); }); app.listen(PORT, () => { console.log(`MCP Server running on port ${PORT}`); }); ``` ### 3. Add Authentication For security, add API key authentication: ```typescript app.use((req, res, next) => { const apiKey = req.headers['x-api-key']; if (apiKey !== process.env.API_KEY) { return res.status(401).json({ error: 'Unauthorized' }); } next(); }); ``` ## Deployment Platforms ### Option 1: Vercel **Pros**: Easy deployment, free tier, global CDN **Cons**: Serverless functions have 10s timeout 1. Install Vercel CLI: ```bash npm i -g vercel ``` 2. Create `vercel.json`: ```json { "version": 2, "builds": [ { "src": "dist/server.js", "use": "@vercel/node" } ], "routes": [ { "src": "/mcp", "dest": "dist/server.js" } ] } ``` 3. Deploy: ```bash vercel ``` ### Option 2: Cloudflare Workers **Pros**: Edge computing, fast, generous free tier **Cons**: Different runtime environment See: https://blog.cloudflare.com/remote-model-context-protocol-servers-mcp/ ### Option 3: Railway **Pros**: Simple, supports long-running processes **Cons**: No free tier 1. Connect GitHub repo 2. Deploy automatically on push ### Option 4: Fly.io **Pros**: Good for WebSocket/SSE, free tier **Cons**: Requires Docker knowledge ## Client Configuration (Remote) Once deployed, clients configure differently: ```json { "mcpServers": { "vegalite": { "type": "sse", "url": "https://your-deployment-url.com/mcp", "headers": { "x-api-key": "your-api-key-here" } } } } ``` ## Considerations ### Performance - Add caching (Redis, Cloudflare KV) - Implement rate limiting - Optimize response sizes ### Security - Use HTTPS only - Implement proper authentication (OAuth, API keys) - Add CORS headers appropriately - Validate all inputs ### Monitoring - Add logging (Winston, Pino) - Track usage metrics - Set up error alerting ### Costs - Vercel: Free tier generous, then $20/month - Cloudflare Workers: Free tier very generous - Railway: ~$5-20/month - Fly.io: Free tier available ## Migration Path 1. **Phase 1**: Keep STDIO for local development ✅ (Current) 2. **Phase 2**: Add HTTP/SSE transport support 3. **Phase 3**: Deploy to staging environment 4. **Phase 4**: Add authentication and monitoring 5. **Phase 5**: Deploy to production 6. **Phase 6**: Support both local and remote ## Example: Dual Transport Support both local and remote: ```typescript // src/index.ts const transport = process.env.REMOTE ? new SSEServerTransport('/mcp', res) : new StdioServerTransport(); await server.connect(transport); ``` ## Testing Remote Deployment Before deploying: 1. Test locally with HTTP transport 2. Test authentication flow 3. Test with actual MCP clients 4. Load test the endpoints 5. Monitor error rates ## References - [MCP Remote Servers Guide](https://modelcontextprotocol.io/docs/concepts/transports) - [Cloudflare MCP Blog Post](https://blog.cloudflare.com/remote-model-context-protocol-servers-mcp/) - [Microsoft Remote MCP Server](https://learn.microsoft.com/api/mcp) --- **Status**: 📝 Planning document - implementation not started **Priority**: Low - Local deployment works well for now **Next Steps**: 1. Finish local version features first 2. Gather user feedback 3. Assess need for remote deployment 4. Implement if there's demand