Authors @BENJAMIN RATIARISOLO @Jean Philippe Orsini
Introduction
In today's AI-driven landscape, developers are constantly seeking ways to enhance their AI assistants with specialized capabilities. While large language models excel at generating content and understanding context, they often lack the structured decision-making capabilities needed for complex business scenarios.
The Model Context Protocol (MCP) is an open standard that enables AI assistants to discover and use external tools and services. By leveraging MCP, AI assistants can extend their capabilities beyond their training data, accessing specialized functions without requiring custom integration code.
With IBM Decision Intelligence (DI), you gain access to powerful tools for authoring, governing, deploying, and executing business decisions. Through its MCP server, you can unlock new possibilities, allowing AI assistants to invoke decisions directly.
This blog post provides step-by-step instructions for connecting Claude with decisions deployed in DI, demonstrating how easily you can extend your AI assistant's capabilities with powerful decision intelligence features. You will learn to deploy the decision sample of loan approval, connect Claude to this decision through configuration, and leverage it directly in your Claude conversations.
Prerequisites
Before starting, make sure you have the following installed or configured:
Step 1 - Deploy The Loan Approval Decision
The following instructions will import the Banking sample in the DI Authoring and deploy the Loan Validation decision service to the DI Runtime
Log in to Decision Intelligence, click on the button "New decision automation", enter a name (ex: LoanValidation), and click on the button "Create"
Click on the button "New decision"
Select the "Industry samples" then "Banking" and click on the button "Import"
Select the "Share changes" panel, and click on the "Share" button.
Select the "Deploy" panel, enter the name of the version (ex: 1.0.0), and click on the "Create" button
Open the "1.0.0" version and click on the button "Deploy"
The decision service "Loan Approval" is now ready to be invoked by any application through its REST API.. including your favorite AI Assistant!
Step 2 - Setting Up the MCP Server in Claude
Locate Your Claude Desktop Configuration File
The configuration file location depends on your operating system:
- macOS:
~/Library/Application\ Support/Claude/claude_desktop_config.json
- Windows:
%APPDATA%\Claude\claude_desktop_config.json
- Linux:
${HOME}/.config/Claude/claude_desktop_config.json
Configure the MCP Server
Add the following configuration to your Claude desktop configuration file:
{
"mcpServers": {
"decision-intelligence": {
"command": "npx",
"args": [
"-y",
"di-mcp-server@latest",
"--apikey",
"<YOUR_API_KEY>",
"--url", "https://<YOUR_TENANT_NAME>.decision-prod-us-south.decision.saas.ibm.com/ads/runtime/api/v1"
]
}
}
Configuration Parameters:
- Replace
<YOUR_API_KEY> with your IBM Decision Intelligence API key
- Replace
<YOUR_TENANT_NAME> with your specific tenant name
Restart Claude Desktop
After saving the configuration file, start the Claude Desktop application to load the new MCP server configuration.
Each decision operation deployed in your IBM Decision Intelligence instance is now available as tools in Claude.
Step 3 - Select the Loan Approval decision operation
Click on the "Search and tools" button to display the list of configured MCP servers, then click on the "decision-intelligence" server.
Disable all tools except "Loan_Approval_approval". The name of the tool is the concatenation of the name of the decision service and the operation.
The experience with Claude is now enriched with the ability to get a loan approval from the decision service of DI!
Step 4 - Conversation with Claude based on the decision
Now, you can create a new chat and ask for the approval of a loan.
Conclusion
By integrating IBM Decision Intelligence with Claude through the DI MCP server, you've created a powerful combination that brings enterprise-grade decision-making capabilities to conversational AI. This integration enables consistent decision making by leveraging tested, governed business logic instead of relying on AI interpretation.
As MCP is an open standard with rapidly growing adoption across the AI ecosystem, this same approach can be replicated with any AI assistant that supports the protocol.