MuleSoft MCP server setup and MCP Inspector testing

Jitendra Bafna demonstrates how to build an MCP server in MuleSoft using Anypoint Studio, then test its tool with MCP Inspector. The example wraps an existing flow that creates Salesforce accounts. That flow maps input fields to the account object and formats the response.

The tutorial adds a flow reference and an MCP server tool listener. Bafna configures an HTTP listener, selects streamable HTTP, and sets the endpoint to /Salesforce/MCP. The demo uses HTTP; he explains that HTTPS requires a TLS context. He assigns the server version 1.0.0 and names the tool create account.

Much of the walkthrough covers the tool description and JSON input schema. Bafna explains how these help an AI agent select a tool and map user input to its fields. In this example, only the account name is required. He configures the response to return the raw payload as text.

Testing requires Node and NPM on the machine. After starting MCP Inspector, Bafna selects streamable HTTP through the proxy and connects to the local server on port 8081. He lists the available tool, enters sample account details, and runs it. The demonstrated response reports success and includes an account ID. Claude and ChatGPT integration remain topics for later videos; this episode stops at testing the server with Inspector.