Skip to content

Turn a REST API into an MCP server in 5 minutes

Updated October 24, 2026 · Published October 24, 2026

To turn a REST API into an MCP server, import its OpenAPI file into Voozeo MCP, check the generated tools, test one in the sandbox and paste the server URL and key into your AI client. This guide shows each step.

Before you start

You need a Voozeo account (free, created in a minute from the sign-up screen), an OpenAPI 3.0 or 3.1 description of your API (JSON or YAML, as a file or at a public HTTPS address) and an AI client that supports MCP. The spec can weigh up to 5 MB. A Swagger 2.0 file is rejected with UNSUPPORTED_SPEC_VERSION: convert it to OpenAPI 3 first.

Here is a minimal spec that produces two tools, listProducts and getOrder:

openapi: 3.0.3
info:
  title: Shop
  version: 1.0.0
servers:
  - url: https://api.example.com
paths:
  /products:
    get:
      operationId: listProducts
      summary: List the products
      responses:
        "200": { description: OK }
  /orders/{id}:
    get:
      operationId: getOrder
      summary: Read an order by its identifier
      parameters:
        - name: id
          in: path
          required: true
          schema: { type: string }
      responses:
        "200": { description: OK }

Step 1: create the server

  1. On "My servers", click "Create a server".
  2. Enter a name and, if you wish, a description, then click "Create".
  3. Copy the MCP key from the "Your MCP key" dialog into a password manager. It is shown only once.

Step 2: import the OpenAPI file

  1. Open the "Sources" tab and click "Import an OpenAPI spec".
  2. Choose "File" and select your file, or choose "URL" and enter the public HTTPS address of the spec.
  3. Click "Import".

A message shows the number of generated tools. The server becomes "active" after the first successful import. Warnings, if any, are listed in the "Sources" tab. See importing an OpenAPI API.

Step 3: check the tools

Open the "Tools" tab. Each operation is a tool that you can enable or disable, rename and describe. Good descriptions help the AI choose the right tool:

  • name with a verb and an object, such as list_orders;
  • say when to use the tool, and when not to;
  • give the units and formats of the parameters.

Your plan sets how many tools can be active per server: 50 on the Free plan. Disable the write and delete tools that your AI does not need. See choosing and describing tools.

Step 4: set the target API

In the "Target API" tab, check the base URL. It comes from the spec; you can enter your own, which a new import never overwrites. Then choose the authentication your API expects: none, API key, bearer token, basic or OAuth 2.0. The secret is stored encrypted and is never shown again. With OAuth 2.0, each person signs in with their own account at your API provider and no MCP key is needed. See target API and authentication.

Step 5: test, then connect

  1. In the "Test" tab, pick a tool, fill in its arguments and click "Run". The result shows the status, the duration and the HTTP code of your API. The sandbox works even before you connect an AI.
  2. In the "Connection" tab, copy the block for your client and paste it where it belongs: Claude, ChatGPT, Cursor or VS Code.
  3. Ask your AI to use the tool. The call appears in the "Log" tab.

What to do next

  • Re-import the spec whenever your API changes: your customizations are kept for the operations that still exist.
  • Keep an eye on usage: the quota is counted in calls per month, and the Free plan includes 1,000. See pricing.
  • Read how to secure an AI's access to your API.
Get started