WatsonX

You set up an MCP server to let Claude Code delegate tasks to watsonx.ai foundation models. This server enables text generation, chat, embeddings, and model listing by routing requests from Claude to watsonx.ai, helping you leverage enterprise-grade IBM models inside your workflow.

How to use

You integrate the watsonx MCP Server with your Claude Code workflow to access watsonx.ai capabilities. Use watsonx_generate for text generation, watsonx_chat for conversational interactions, watsonx_embeddings to create embeddings, and watsonx_list_models to see available models. When Claude identifies the need for IBM-specific models or embedding generation, it will delegate those tasks to watsonx.ai through the MCP server.

How to install

{
  "mcpServers": {
    "watsonx": {
      "type": "stdio",
      "command": "node",
      "args": ["/Users/matthewkarsten/watsonx-mcp-server/index.js"],
      "env": {
        "WATSONX_API_KEY": "your-api-key",
        "WATSONX_URL": "https://us-south.ml.cloud.ibm.com",
        "WATSONX_SPACE_ID": "your-deployment-space-id"
      }
    }
  }
}

Configuration and usage notes

Prerequisites you need before starting: a working Node.js environment and npm installed on your system.

Install dependencies for the MCP server:

c d ~/watsonx-mcp-server
npm install

Add the MCP server definition to Claude Code with the following configuration snippet, which wires the watsonx MCP server as a stdio process that runs index.js from your local path and passes required environment variables for authentication and routing to watsonx.ai.

{
  "mcpServers": {
    "watsonx": {
      "type": "stdio",
      "command": "node",
      "args": ["/Users/matthewkarsten/watsonx-mcp-server/index.js"],
      "env": {
        "WATSONX_API_KEY": "your-api-key",
        "WATSONX_URL": "https://us-south.ml.cloud.ibm.com",
        "WATSONX_SPACE_ID": "your-deployment-space-id"
      }
    }
  }
}

Notes on environment and usage

This MCP server uses the watsonx.ai Studio service in the us-south region with a Lite plan. You should create your own watsonx.ai project and deployment space to align with your IBM Cloud setup.

Two-Agent System overview

You can configure Claude Code (Opus 4.5) to delegate specific workloads to watsonx.ai when IBM-specific capabilities or embeddings are needed. This two-agent arrangement helps you route workload efficiently between Claude and watsonx.ai.

IBM Cloud Resources

Key context: service is watsonx.ai Studio (data-science-experience), Lite plan, region us-south. Create your own watsonx.ai project and deployment space in IBM Cloud.

Integration with other MCP servers

This watsonx MCP Server can operate alongside other MCP servers, enabling a multi-service pipeline where Claude delegates tasks to different backends as needed.

Examples of usage within your workflow

Example: request text generation or embeddings from Claude to feed a retrieval-augmented generation (RAG) setup using watsonx models.

Available tools

watsonx_generate

Generate text using watsonx.ai models

watsonx_chat

Chat with watsonx.ai chat models

watsonx_embeddings

Generate text embeddings for semantic search or RAG pipelines

watsonx_list_models

List available watsonx.ai foundation models