Using watsonx.ai Flows Engine with Model Context Protocol (MCP)

Here's a step-by-step tutorial for setting up and deploying a project with wxflows, including installing necessary tools, deploying the app, and running it locally.

This example consists of the following pieces:

MCP TypeScript SDK (mcp server)
wxflows SDK (tools)

You can use any of the supported MCP clients.

This guide will walk you through installing the wxflows CLI, initializing and deploying a project, and running the application locally. We’ll use google_books and wikipedia tools as examples for tool calling with wxflows.

Before you start

Clone this repository and open the right directory:

git clone https://github.com/IBM/wxflows.git
cd examples/mcp/javascript

Step 1: Set up wxflows

Before you can start building AI applications using watsonx.ai Flows Engine:

Sign up for a free account
Download & install the Node.js CLI
Authenticate your account

Step 2: Deploy a Flows Engine project

Move into the wxflows directory:

cd wxflows

There's already a wxflows project for you set up this repository with the following values:

Defines an endpoint api/mcp-example for the project.
Imports google_books tool with a description for searching books and specifying fields books|book.
Imports wikipedia tool with a description for Wikipedia searches and specifying fields search|page.

You can deploy this tool configuration to a Flows Engine endpoint by running:

wxflows deploy

This command deploys the endpoint and tools defined, these will be used by the wxflows SDK in your application.

Step 3: Set Up Environment Variables

From the project’s root directory copy the sample environment file to create your .env file:

cp .env.sample .env

Edit the .env file and add your credentials, such as API keys and other required environment variables. Ensure the credentials are correct to allow the tools to authenticate and interact with external services.

Step 4: Install Dependencies in the Application

To run the application you need to install the necessary dependencies:

npm i

This command installs all required packages, including the @wxflows/sdk package and any dependencies specified in the project.

Step 5: Build the MCP server

Build the server by running:

npm run build

Step 6: Use in a MCP client

Finally, you can use the MCP server in a client. To use with Claude Desktop, add the server config:

On MacOS: ~/Library/Application Support/Claude/claude_desktop_config.jsonOn Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "wxflows-server": {
      "command": "node",
      "args": ["/path/to/wxflows-server/build/index.js"],
      "env": {
        "WXFLOWS_APIKEY": "YOUR_WXFLOWS_APIKEY",
        "WXFLOWS_ENDPOINT": "YOUR_WXFLOWS_ENDPOINT"
      }
    }
  }
}

You can now open Claude Desktop and should be seeing the tools from the wxflows-server listed. You can now test the google_books and wikipedia tools through Claude Desktop.

Summary

You’ve now successfully set up, deployed, and run a wxflows project with google_books and wikipedia tools. This setup provides a flexible environment to leverage external tools for data retrieval, allowing you to further build and expand your app with wxflows. See the instructions in tools to add more tools or create your own tools from Databases, NoSQL, REST or GraphQL APIs.

Support

Please reach out to us on Discord if you have any questions or want to share feedback. We'd love to hear from you!

Installation

To use with Claude Desktop, add the server config:

On MacOS: ~/Library/Application Support/Claude/claude_desktop_config.jsonOn Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "weather-server": {
      "command": "/path/to/weather-server/build/index.js"
    }
  }
}

Debugging

Since MCP servers communicate over stdio, debugging can be challenging. We recommend using the MCP Inspector, which is available as a package script:

npm run inspector

The Inspector will provide a URL to access debugging tools in your browser.

Using watsonx.ai Flows Engine with Model Context Protocol (MCP)