ibm-decision-intelligence

This MCP server exposes IBM Decision Intelligence and IBM Automation Decision Services decision runtimes as MCP tools that can be invoked by supported AI assistants and MCP clients. By running the server with your Decision Intelligence API key, you enable your client applications to query and expose individual decision services as MCP tools, allowing orchestration agents to request decisions via MCP transports (stdio or HTTP) and receive structured decision results. The server supports both Decision Intelligence (DI) and Automation Decision Services (ADS) runtimes, so you can surface decision services deployed in either platform as reusable MCP tools. To use it, start the MCP server with your DI runtime base URL and credentials, then point your MCP client to the exposed tools; each decision service deployed in the runtime can be surfaced as a separate tool for consumption by AI assistants and orchestration workflows.

Prerequisites:

• Node.js (LTS) and npm installed on your machine
• Access to your IBM Decision Intelligence runtime (DI) or IBM Automation Decision Services (ADS) runtime URL

Install and run (example using npx):

1. Ensure you are logged in to npm if required by your environment.
2. Run the MCP server via npx (no global install required):

npx -y di-mcp-server --di-apikey <YOUR_DI_API_KEY> --url https://mytenant.decision-prod-us-south.decision.saas.ibm.com/ads/runtime/api/v1

This command starts the MCP server and exposes each deployed decision service as an MCP tool for your AI assistants to consume. If you prefer to configure with a specific environment or deployment, adjust the --url and authentication options accordingly.

Optional authentication modes for ADS (Zen or Basic) can be used instead of DI API key by following the CLI options documented in the README of the project. See developing and contributing guides for more details.

Tips and common issues:

• Ensure the DI/ADS runtime URL is reachable from the machine running the MCP server.
• When using npx, the server runs with the last deployed version of the di-mcp-server package; pinning a version is possible by replacing di-mcp-server with di-mcp-server@<version> if needed.
• If you rotate API keys, update the invocation command to use the new key.
• The MCP server supports both STDIO and streamable HTTP transports; configure your MCP client accordingly if you need streaming capabilities.
• Environment variables like DI_APIKEY can be documented for your deployment environment to avoid hard-coding credentials in scripts.