MCP server development
MCP servers that give AI tools access to your product
The Model Context Protocol lets Claude, ChatGPT, Cursor and VS Code call real tools. I build MCP servers for products that want AI agents as users, and for teams that want their assistants to reach internal systems, with the authentication and limits a public server needs.
Is this for you
A good fit if
- You run a SaaS or API and want AI agents and assistants to use it directly.
- Your team wants Claude or Cursor to query internal data or trigger internal jobs safely.
- You have an MCP prototype and need it production-ready: auth, scopes, tests and deployment.
What I build
What this looks like in practice
Product MCP servers
A remote MCP server in front of your API, with tools designed for how agents actually call them, typed schemas and clear errors.
Internal tool servers
Servers that expose databases, reports or jobs to your team's assistants with read-only roles and an audit log.
Auth and safety
OAuth and scoped tokens, rate limits, input validation and protection against prompt injection through tool output.
Testing and deployment
Automated tests against the protocol, stateless deployment on your infrastructure, and monitoring.
How it works
From first call to handover
A short call
You describe the problem, the systems you already have and the deadline. I ask the questions that decide the shape of the build.
A written scope
You get a plain document with what I'll build, what I won't, the milestones and the estimate, before any work starts.
Weekly progress
I build in small, working increments you can try, with a short update each week. Nothing disappears for a month.
Handover
Code in your repository, deployed on infrastructure you control, with documentation your team can run it from.
What you get
At the end
- A deployed MCP server your users or team can connect to
- Tool schemas and docs for each tool
- Auth, scopes and rate limits configured
- Tests that run in CI
Stack
Usually built with
- Model Context Protocol
- TypeScript
- Next.js
- NestJS
- OAuth
- Zod
Related work
Things I've built along these lines
Open SourceBenchlineBenchline is a free, open-source directory of browser tools for developers, SEO, PDFs and images. Every tool is also callable by AI agents in Cursor, Claude or VS Code through a stateless MCP server, with no API key.
SaaSReelspanReelspan is a social publishing API and dashboard that posts one video, photo set or text post to YouTube, TikTok, Instagram, X, LinkedIn and more at once.
Open Sourceagent-rulesOpen-source, modular instructions for AI coding agents. One source of rules renders into AGENTS.md, CLAUDE.md, Cursor rules, GitHub Copilot instructions, Claude Code skills and Windsurf/Devin formats through a zero-dependency CLI.
How I approach it
Guides I've written on this
- How to Build an MCP Server in Next.js: Streamable HTTP, Auth and Tools10 min read
- How to Build an MCP Server in NestJS: Streamable HTTP, Guards and DI (Tested)7 min read
- Stateless MCP Servers: What the 2026-07-28 Spec Changed and How to Migrate11 min read
- MCP Security in 2026: The OAuth SDK CVEs, Tested, and How to Lock Down Your Server11 min read
- How to Test an MCP Server: Vitest, In-Memory Transport, Streamable HTTP and the Inspector CLI5 min read
- MCP vs API: When You Need Which, With Real Benchmarks11 min read
Tell me what you need built
Send the problem and a rough timeline. I reply personally, usually within a day, with questions or a next step.