Create an MCP Server with the Postman AI Tool Builder

With Postman and the Model Context Protocol (MCP), you can create an MCP server that provides a standard, consistent two-way communication interface between AI systems and API services. You can expose tools or data through this interface, and an MCP client can then interact with them.

Create the MCP server

To generate an MCP server, do the following:

  1. Go to the AI Tool Builder.
  2. Search for an API in the Postman API Network and click it.
  3. Click the API requests you want to expose as tools in your MCP server.
  4. Click Add Requests To Agent.
  5. Click Generate.
  6. Postman generates the files for your server and compresses them in a zip file. When the file is ready, click Download ZIP. By default, the zip file is named postman-mcp-server.
  7. Extract the zip file to create a directory for your MCP server.

Run the MCP server locally

Your generated MCP server code is clean, well-documented TypeScript that you can edit and extend. Before running the server locally, make sure you have the following required tools:

To start the server, do the following:

  1. Open a terminal and navigate to the directory where you extracted the zip file.

  2. Enter npm install.

  3. Enter node mcpServer.js. Your MCP server starts running.

Send a request to the MCP server

To send requests to the server while it's running locally, do the following:

  1. Open a workspace in Postman and click New > MCP icon MCP.

  2. Enter node path-to-server/mcpServer.js, replacing path-to-server with the location where you unarchived the server zip file.

  3. Click Connect.

  4. Click Grant Access.

  5. Click one of the MCP server's tools, then click Run. The response appears in the Response pane.

To learn more about MCP requests, see Create an MCP request for your collection. To learn more about extending the server, deploying it in Docker, or running it from Claude Desktop, see the README in the extracted zip file.

Last modified: 2025/05/15