Configuring or using custom own LLM models in Pega Agents

Hello Everyone,

Is there any possibilities or there is any configuration feasibility to allow clients to select and use their own LLM models within AI Agents, instead of relying solely on the models provided by Pega?

I am looking for a solution where we can use Pega Agent rule as it is with all its feasibility but it should work with the custom own LLM models, where our client can select their own model and use the Pega Agent capabilities on their business process / Life cycle instead of fully rely on Pega provided models, till now I haven’t get any documentation or solution available to achieve this requirement (specifically for Pega Agent). Is this current limitation ? Anything coming in upcoming versions?

Please let me know if anyone have any idea or tried before to achieve this.

In Pega Infinity 26, you can use your own LLM through Pega’s Bring Your Own Model (BYOM) capability. First, expose your custom LLM through a secure, OpenAI-compatible HTTPS API. Share the model endpoint, model name, and authentication details with Pega Support to configure it through the Pega GenAI Gateway. This allows supported GenAI requests to be routed to your custom model instead of Pega-managed models. You can retain your existing Pega Agent rules, workflows, and business logic, although direct replacement of an Agent rule’s underlying model should be confirmed for your deployment

Hi @ArijitG96, this goes beyond simple rule configuration and is a question of deployment. As such it depends on whether you are on an Infinity 26 and if you are on a Pega Cloud, or a Client-managed setup.

A good place to start would be this document - Integrating with Pega GenAI
Note: you can also switch the platform version, but own LLMs are supported only starting from Infinity 26, as far as I know.

In short you need to choose which deployment scenario to use depending on the following situations:

GenAI Own LLMs
Pega Cloud Out-of-the-box PremBridge Extend
Client-managed PremBridge Pega-Managed PremBridge Local

Then follow the corresponding deployment instructions in the article.

You can find instructions for Prembridge Local here