What is IBM wxflows?

IBM wxflows is a tool platform developed by IBM that enables users to build, test, and deploy tools for any data source, particularly leveraging the Model Context Protocol (MCP) for seamless integration with AI models like Claude Desktop.

How to use IBM wxflows?

To use IBM wxflows, follow these steps:

Clone the repository and navigate to the project directory.
Sign up for a free IBM wxflows account and install the CLI.
Deploy the project using the wxflows deploy command.
Set up environment variables by copying the .env.sample file to .env and adding your credentials.
Install dependencies with npm i and build the MCP server with npm run build.
Configure the MCP server in your client (e.g., Claude Desktop).
Run and test the tools like google_books and wikipedia in your application.

Key features of IBM wxflows

Flexible tool deployment: Build and deploy tools for any data source.
Model Context Protocol (MCP) integration: Seamlessly connect tools with AI models.
Pre-built tools: Includes tools like google_books and wikipedia for data retrieval.
Extensible: Add or create custom tools from databases, NoSQL, REST, or GraphQL APIs.

Use cases of IBM wxflows

Data retrieval and analysis: Use tools like google_books and wikipedia to fetch and process data.
AI model enhancement: Integrate external tools with AI models (e.g., Claude Desktop) to extend their functionality.
Custom tool development: Build and deploy custom tools for specialized data sources.
AI workflow automation: Automate data retrieval and processing workflows using wxflows.

FAQ from IBM wxflows

What is MCP?

MCP (Model Context Protocol) is a standard for integrating external tools with AI models.
How do I deploy a wxflows project?

Use the wxflows deploy command in your project directory.
What tools are included in wxflows?

The example includes google_books and wikipedia tools, but you can add more or create your own.
Where can I get support?

Join the IBM wxflows Discord community for help and feedback.

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.json On 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: