MCP servers for your product's API.
I build Model Context Protocol servers that let AI assistants and agents use your product. They run on the same operations as your REST API, so an agent can do what a user can, with the same permissions and the same tests behind it.
I reply personally within 2 business days. Rates come up on a call, once the scope is clear.
What I build
One registry for REST and MCP
Each operation is defined once and served both ways, so the API and the agent tools can't drift apart.
Tools an agent can choose between
Clear names, short descriptions and schemas loaded only when a tool is used, so a large API doesn't fill the model's context before it starts.
Your users' permissions
The agent acts as a user, with that user's access, and every call is checked on the server like any other request.
Tests on every operation
Failing test first for the API and the MCP surface alike, so a change that breaks an agent fails before it ships.
A good fit
- A SaaS or platform whose customers want to use it from Claude, ChatGPT or Cursor
- A team that wants agents working inside its own product, with real permissions
- An MCP server that loads too many tools or breaks when the API changes
Not a fit
- A one-off demo with no product or API behind it
- Research on new models, which needs an ML specialist
How it runs
- 01
You write to me
Tell me what you're building through the form. I reply personally within 2 business days.
- 02
We talk it through
A call on the problem, the fit and what it should cost. Rates come up here, once the scope is clear.
- 03
I scope it in writing
What ships first, what it's built on and how it's tested, agreed before the build starts.
- 04
I build it test-first
Each change starts with a failing test, and you see the product running as it grows.
- 05
It's yours to run
The code, the tests and the notes your team needs to keep building on it.
Proof you can check
- Case studySharpOS220 operations in one registry, every one served over REST and MCP, about 130 MCP tools. The platform my studio runs on.
- Case studyRelayMy own AI harness, where four chat agents work in a project folder through 73 MCP tools.
- PostToo many MCP tools?How one registry and lazy schemas serve 130 tools for about half the context a strict mirror costs.
Questions, answered
Do I need an MCP server if I already have a REST API?
Not always. An agent can call a REST API through a generic tool, but it has to guess which endpoint does what. An MCP server describes your operations as tools the model can pick and call safely. If your customers ask to use your product from an AI assistant, you need one.
Which assistants can use it?
Any MCP client: Claude, ChatGPT, Cursor and the coding agents that support the protocol. The server is the same for all of them.
How many MCP tools is too many?
There's no fixed number. What hurts is loading every tool's schema before the model starts. With lazy schemas, SharpOS serves about 130 tools for half the context a strict mirror would take.
Can you add MCP to a product you didn't build?
Yes, if it has an API or a codebase I can work in. I start from the operations you already have, then decide which become tools.
What does it cost?
I don't publish rates. I quote once I know the scope, in our first call.
Tell me what you're building.
One message is enough to start. I read every one myself and reply within 2 business days.
Need a website for a small business? That's my studio, SharpHaw (opens in a new tab).