Make Me An MCP
[ MAKE ME AN MCP ]FREE ALPHAby Think North

Make your API callable by Claude, Codex & ChatGPT — without building an MCP.

Point our MCP at your codebase and it reads your routes, proposes the tools an agent should see, and stands up a live MCP server — behind keys, access scopes and rate limits. No gateway to hand-build, no manifest to hand-write. Prefer a spec? Import your OpenAPI and you're live in minutes either way.

// free while we're in alpha · join in and share your feedback · no pricing, no sales call — you help shape it
in Claude / Cursor / Codex · our MCP builds yours
# you, to the agent "Set up an MCP for the API in this repo." # the agent calls OUR MCP's tools › make-me-an-mcp · analyze_repo() ↳ 12 routes · 3 capability groups · OpenAPI found › make-me-an-mcp · propose_tools() ↳ Refunds · Charges · Customers — with when-to-use hints › make-me-an-mcp · create_server({ confirm: true }) ↳ live › /mcp/s/acme-payments/ ↳ keys · rate limits · access scopes ready # point any agent at that URL — done.
Two ways in. Both end with a live server. // the whole loop — your code to a URL any agent can call
  1. 1Connect our MCP in your agent — or import an OpenAPI spec
  2. 2It reads your code & proposes the tools
  3. 3Confirm groups, scopes & rate limits
  4. 4Mint a key
  5. 5Point Claude/Codex/ChatGPT at /mcp/s/<slug>/
[ 001 ]
The gap

Your API works. Agents still can't use it.

{ }

MCPs are hand-built

Wrapping each route as an agent tool — schemas, auth, descriptions — is fiddly, repetitive work that goes stale the moment the API changes.

?

Agents guess blindly

With no clear description of what each endpoint returns, an agent can't reliably decide which one to call — or what to do with the response.

No controls in front

Opening an API to agents means keys, scopes, rate limits and quotas. Standing that gateway up is its own project.

[ 002 ]
Why our way

The only MCP builder that reads your code for you.

Most tools make you describe every endpoint by hand, then drift from the real API. Make Me An MCP has its own MCP. You connect it inside Claude, Cursor or Codex and ask it to set up your server. It reads your routes and OpenAPI, proposes the capability groups and when-to-use hints, and scaffolds a live server you confirm — so the manifest agents read is generated from your actual code, not copy-pasted prose.

The old way

  • Hand-write an MCP server per API
  • Re-describe every endpoint and schema by hand
  • Copy tool descriptions into every agent
  • Bolt on a separate gateway for keys & limits
  • Watch it drift the next time the API changes

With Make Me An MCP

  • Our MCP reads your code and proposes the tools
  • Output signatures read from real responses
  • One manifest, generated — every agent reads it
  • Keys, scopes, rate limits & quotas built in
  • Regenerate from the source whenever it changes
[ 003 ]
What you get

An MCP control plane for your API.

core

Our MCP builds yours

Connect Make Me An MCP in your agent and ask it to set up your server. It reads your routes, proposes the tools, and scaffolds a live MCP — guided end to end.

import

OpenAPI in one paste

Already have a spec? Paste or link your OpenAPI/Swagger and every path becomes a grouped, agent-ready tool automatically.

discovery

Output-signature reading

It reads the shape of each endpoint's response so the manifest tells agents exactly what comes back — not just what to send.

agent-facing

Editable AI "yellow pages"

An auto-drafted manifest you can edit — the description agents read to choose which API to invoke, and why.

controls

Keys, scopes & rate limits

Internal, enterprise, or private. Every consumer connects with a key scoped to what it's allowed to touch, with recommended rate limits and quotas set for you.

visibility

Live usage & metering

Every call is metered per key and endpoint, with rate-limit and quota hits tracked — so you can see exactly what agents are doing with your API.

0 servers
to build by hand
1 manifest
agents read to decide
3 agents
Claude · Codex · ChatGPT
keys
scoped, rate-limited, metered
[ 004 ]
Who it's for

Built for teams whose API should be doing more.

A good fit

  • Platforms that want agents and partners building on their API
  • Teams exposing internal tools to their own AI agents
  • API products that need keys, scopes and rate limits in front
  • Anyone who'd rather not hand-write and maintain an MCP server

What it replaces

  • Hand-rolled MCP servers that drift from the real API
  • Tool descriptions copy-pasted into every agent
  • A separate gateway for keys, rate limits and metering
  • Docs that humans read but agents can't act on
[ 005 ]
Questions

The short version.

What does it cost?

It's a free alpha right now. Join in, build your MCP, and help us shape the product — and its pricing — before we set any. Your feedback is the point.

How does it read my code?

Make Me An MCP has its own MCP. Connect it in Claude, Cursor or Codex and ask it to set up your server. It reads your routes and OpenAPI, proposes the tools and groups, and scaffolds a live server you confirm.

Do I have to rewrite my API?

No. It builds the agent-facing surface on top of the API you already run — reading each endpoint's output signature and generating the manifest for you.

How do I control who uses it?

Every consumer connects with a key scoped to what it's allowed to touch. You decide whether a server is internal, enterprise, or private, with recommended rate limits and quotas set for you.

How do agents decide which API to call?

They read the manifest — the auto-drafted, editable "yellow pages" describing each group and endpoint, when to use it, and the shape of what it returns.

Join the free alpha

Tell us about your API and we'll get you set up — the code-reading MCP, the import, the controls. In return, we want your honest feedback while we build this with you.

No spam. Free during alpha. Already have access? open the console →