Skip to content
MCP Integration Server live preview
All projects

MCP Integration Server

Let an AI agent operate across 15+ services from one workflow.

AI assistants could see infrastructure but not act on it — each service (Cloudflare, Vercel, Notion, Strava…) needed its own auth and API shape, so agents stayed read-only.

Tech Domain · Book Dev|Published|Evidence: Live|TypeScript, MCP, Node.js, REST APIs

A public deployment is available to try.

Case Study

Before

AI assistants needed access to real infrastructure (deployments, secrets, analytics, calendars) but each service had different APIs and auth mechanisms. Integrating 15+ services manually was unsustainable.

After

AI assistants can now manage deployments, rotate secrets, check analytics, create calendar events, and query fitness data across 15+ services through a standardized MCP interface, instead of hopping between separate consoles and auth flows for each one.

How it's solved

Built a custom MCP server with 100+ tools spanning Cloudflare Workers, Vercel deployments, Infisical secrets, Google Workspace, Strava fitness data, Notion pages, and more. Each tool follows the MCP protocol with proper input schemas and error handling.

Trade-offs

Tool coverage tracks whichever services this operation actually uses — not a general-purpose MCP server for arbitrary stacks.

Related Articles

Further reading on related topics from the blog

Related Projects

More in Tech Domain · Book Dev

MCP Integration Server - Chaowalit Greepoke Portfolio