Exposing Pega Workflows to LLMs via MCP

General
1

Hi everyone,

I’m exploring how to integrate Pega workflows with (LLMs) using the (MCP) and would appreciate guidance from anyone who has worked on a similar architecture.

My goal is to expose Pega business capabilities as MCP tools so that an LLM (such as ChatGPT or Claude) can orchestrate Pega workflows through natural language.

For example:

  1. A user says, “I want to apply for a personal loan.”
  2. The LLM invokes an MCP tool like createLoanCase().
  3. Pega creates the case and returns the required/missing fields (loan amount, income, PAN, etc.).
  4. The LLM collects the missing information conversationally.
  5. The LLM calls another MCP tool to update the case.
  6. Pega advances the case through stages, performs validations, and invokes integrations with external systems.

Thanks in advance..

2

Hi Avinash,

Pega has actually introduced a native solution for this requirement. Starting from Infinity 25.1.3, there is a new rule type called Service MCP.

Instead of you needing to build and host a custom external MCP server from scratch to wrap Pega’s REST or DX APIs, this rule type allows Pega to natively act as an MCP server.

For detailed documentation about it, we may need to wait until release of Infinity 26.

3

Please find the steps

  1. Create the Service MCP in 25.1.3 or above
  2. Enable MCP Service for your application.
  3. MCP URL will be provided once enabled.
  4. Configure the MCP URL in Claude → Settings → Connectors → Customize
  5. Configure the OAuth 2.0 Authentication Type (Client Id and secret )
  6. Test your changes
4

This article shows the configuration steps in Infinity '26. This one uses Claude as LLM to call Pega MCP tool to create a case in Pega. How to expose Pega as MCP server for external AI clients (Infinity '26)