Why I Built the Universal GM Operating Framework

When I first started experimenting with AI Game Masters, I kept running into the same problems many people experience:

• inconsistent rulings

• forgotten dice rolls

• the AI inventing new rules

• narration and mechanics blending together

• NPCs acting without motive

• and campaigns that slowly drifted away from the system’s RAW

I wasn’t trying to replace human GMs.
I wanted AI to behave like a competent, reliable referee — something you could use for solo campaigns, encounter prototyping, or worldbuilding without the experience breaking down.

So I started designing a structured solution.

The more I worked on it, the more I realised the issue wasn’t with the TTRPG systems themselves — it was with the lack of a controlling framework that told the AI how to think, when to roll, why to act, and what rules to prioritize.

That idea became the Universal GM Operating Framework.

My goals for the Framework are simple:

• Keep the AI consistent over long campaigns

• Make RAW matter again, even when using a language model

• Give players confidence that the AI won’t derail the system

• Support solo players, designers, and GMs who want a reliable co-GM

• Reduce cognitive load, so the AI handles rules while the human focuses on story

• Allow any system to plug into it, from D&D to Marvel Multiverse to homebrew systems

• Promote better AI behaviour, not just better prompts

I hope this Framework gives people a way to explore their favourite systems with an AI that respects the mechanics, the theme, and the world they’re playing in.
I also hope it encourages more experimentation in how we blend TTRPG structure with LLM reasoning.

This isn’t meant to replace GMs — it’s meant to be a tool that expands what’s possible.

If this Framework helps even a few players run smoother solo sessions, prototype encounters, or design better systems, then it’s achieved what I built it for.