LABSCO SUMMARY

Published-versus-draft versioning is the feature that separates this from a wrapper script.

Anyone can proxy REST to MCP. What breaks in practice is editing a live tool definition and discovering that production agents changed behaviour mid-conversation. Keeping production on immutable published versions while edits land in a draft — and giving agents a separate editor endpoint to work on the draft — is the difference between a demo and something you can operate. The per-call price on the pay-as-you-go plan makes the cost of an agent's retry loop visible, which is worth knowing before you turn one loose.

What it is

A platform for wrapping your own REST API as an MCP server. Each tool maps one-to-one to an endpoint — get_weather calls GET /api/weather — with authentication, headers and body configured per tool. The endpoint address is per customer, so there is no shared URL to connect to.

What you get

Tool definitions mapped to your endpoints, with templated headers passed automatically on every call
Optional MCP Prompts for clients that expect reusable presets
Dev and prod environments with immutable published versions alongside a workspace draft
An editor endpoint so an agent in Cursor can compare published against draft, edit draft tools, and save or publish a version
Tool tracing and audit trails over the calls agents make

Requirements

An account and your own API. Stream is the recommended transport, SSE the alternative; clients without remote support go through mcp-remote with the headers passed as flags. A self-hosted edition runs on your own VM under Docker Compose for data residency.

Setup effort

Get your endpoint, then paste it — add your endpoint to your client with your key

See installation options

USE CASES

01

Exposing an API you already have

One tool per endpoint, no backend changes.

02

Editing tools without breaking production

Work on the draft; production stays on the published version until you promote it.

03

Seeing what agents actually called

Tracing and audit trails over every tool call.

04

Keeping data in your own estate

The self-hosted edition on your own VM.

BEFORE YOU INSTALL

What to know first

Requires

A Rapid MCP server account
An AI agent client (e.g. Cursor, Windsurf, Claude Desktop, OpenAI) with MCP support
A REST API endpoint you want to transform into an MCP tool (e.g. Weather API)

Where it runs

Both documented setups connect to a hosted endpoint.

INSTALLATION

Connect your client

https://rapid-mcp.com/mcp/<YOUR_SERVER_ID>/stream (per customer)

MCP Clients

Any MCP client — URL

ANY MCP CLIENT — URL · JSON

{
  "mcpServers": {
    "RapidMCP": {
      "url": "https://rapid-mcp.com/mcp/<YOUR_SERVER_ID>/stream",
      "headers": {
        "Authorization": "Bearer <YOUR_TOKEN>"
      }
    }
  }
}

<YOUR_SERVER_ID> is your Rapid MCP server's ID, and <YOUR_TOKEN> is your Rapid MCP JWT token. Any header you put here is passed automatically to every tool call.

Any MCP client — npx

ANY MCP CLIENT — NPX · JSON

{
  "mcpServers": {
    "RapidMCP": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://rapid-mcp.com/mcp/<YOUR_SERVER_ID>/stream",
        "--header",
        "api-key:${AUTH_HEADER}",
        "--header",
        "subdomain:${SUBDOMAIN}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer <YOUR_TOKEN>",
        "SUBDOMAIN": "subdomain-v1-example"
      }
    }
  }
}

api-key and subdomain are example headers: Rapid MCP passes them to every tool call, and your tool definitions read them as {{api-key}} and {{subdomain}}. Send the headers your own tools reference, with your own values. api-key must be lowercase and contain no spaces so it maps to {{api-key}} in tool templates.

Values written as ${…} are filled in only by clients that expand variables.

RapidMCP API Converter