Solution Builders have been using AI coding agents for a while now. Anthropic Claude Code, GitHub Copilot, and OpenAI Codex are already part of how many teams write software, and the habit is established: you stay in your editor, you describe what you want, and the agent does the typing. What those agents have not been able to do is build an enterprise application on Pega Platform™. They have no knowledge of the application they are working in, and nothing they produce is subject to the review an enterprise application requires.
Pega Platform extends those agents rather than replacing them. A plug-in connects the agent to the Pega Infinity™ Model Context Protocol (MCP) server, and Pega Infinity skills instruct the agent how to work with Pega Platform. The result is that the agent builds with application context and produces governed Rules — from inside the environment the Solution Builder already uses.
What it means
Two pieces make this work, and it is worth separating them.
- The plug-in connects the AI coding agent to the Pega Infinity MCP server. That connection is what gives the agent access to the application: it is no longer guessing at structure from a prompt, it is working against the real thing.
- The Pega Infinity skills instruct the agent how to work with Pega Platform. This is the part that matters for output quality. The skills tell the agent what a Rule is, where authoring belongs, and how a change is supposed to be made — so the agent builds with application context rather than producing something that merely looks plausible.
What comes out the other end is the detail that counts. The agent authors governed Rules — the same configurable building blocks the platform uses everywhere else — not application-specific code. That is what makes the work reviewable: it is the same kind of Rule a developer would have created through a form, and it is subject to the same review and the same guardrails as everything else in the application.
What it does
Three things follow from extending an existing agent instead of asking teams to adopt a new one.
-
Work in your preferred environment. Solution Builders keep using the AI coding agent and the development environment they already know, and still build on Pega Platform. There is no switch to make and no second toolchain to learn.
-
Human review is built in. Pega Infinity skills include Rules that keep authoring in Branch workspaces, and users explicitly approve every change. Review is part of how the capability operates, not a process layered on top of it.
-
Extend AI from answers to action. An agent that could previously explain or suggest now authors Rules in a governed way — the step from assistance to contribution.
How it works
You install the plug-in, which connects your AI coding agent to the Pega Infinity MCP server. Pega Infinity skills then instruct the agent how to work with Pega Platform, and you describe what the application needs in the environment you normally work in. The agent authors the corresponding Rules.
Authoring stays in Branch workspaces — that constraint comes from the skills themselves rather than from a setting someone has to remember to switch on. Every change is explicitly approved by a person before it goes anywhere, so nothing the agent produces reaches a shared application until someone has looked at it.
Two practical points are worth knowing. The plug-in is available from Pega’s public GitHub repository, so getting started is a matter of installing it against an agent you already run. And MCP connections share the same scope and limitations as Pega AI Assistant — if you know what the Assistant can and cannot touch, you know the boundaries here too.
Availability is less restrictive than a capability like this usually is. External AI authoring is supported from Pega Infinity 24.2.4, and on Pega Infinity 25 and 26, so organizations on any of those releases can start without upgrading first.
A short walkthrough is available in the Pega Academy Topic linked below, in which Ryan Easley, Fellow Specialist Solutions Consultant, shows how an external AI agent authors governed Rules from within the environment Solution Builders already use. It runs 6:25.
Watch this video on Agentic authoring through external agents
