Skip to main content

Generate MCP Server for Your API

liblab lets you generate and host a fully functional MCP (Model Context Protocol) server directly from your OpenAPI spec or Postman collection. This cloud-hosted service provides a public MCP endpoint that’s ready to integrate with AI tools like Claude, Cursor, Windsurf, or GitHub Copilot, without requiring deployment or infrastructure setup.

Unlike the local CLI option, the hosted solution uses liblab's infrastructure to instantly create and deploy your MCP server, returning a remote URL that works with any MCP-compatible client.

The hosted MCP server generated by liblab:

  • Automatically defines the MCP tools based on your API spec.
  • Makes them discoverable and usable by AI assistants like Claude or Cursor.
  • Acts as a secure proxy layer between AI agents and your API.

This makes it easy to expose your API to AI clients while maintaining full control and data privacy. It's important to notice that liblab does not stores logs or call data information.

Before You Start​

To generate an MCP server with liblab, you need:

  • A liblab account.
  • Your API’s OpenAPI spec, either in YAML or JSON format.

If you don’t have an OpenAPI spec yet, you can follow our framework guides to generate one. Alternatively, you can experiment with public specifications such as:

How to Generate and Host an MCP Server with liblab​

  1. Go to the liblab dashboard.

  2. Open the MCP tab and click the MCP Build button to start the process.

  3. Upload your OpenAPI spec file.

    info

    liblab uses your spec to generate the MCP tools. The number and structure of tools depend on the endpoints present in your OpenAPI spec.

  4. Select the authentication method your API uses. This determines how AI clients will authenticate with your MCP server:

    • API Key
    • Basic Auth
    • Bearer Token
    • None

  5. Enter the required authentication details and click Generate MCP Server.

Once the generation process is complete, liblab will display your MCP server URL.

Generated MCP URL

This is the endpoint you'll share with AI tools to enable natural language interaction with your API. To recover the URL later, you can always return to the MCP tab in your liblab dashboard.

Please note that each liblab organization can host only one MCP server at a time. If another team member generates a new one, it will replace the existing deployment.

liblab’s hosted MCP service is billed based on usage. There are no upfront fees, you only pay for what you use.

Connect Your Hosted MCP Server to AI Tools​

Once your server is live, you're ready to integrate it with various tools that support the Model Context Protocol:

Last updated on