The Sable Protocol

Every table needs something that talks back and shouldn't be trusted.

The Sable Protocol is a system-neutral framework for running a live AI language model as an oracle prop at your tabletop game — a laptop, a sentient sword, a dying ship's AI core, anything with a voice and a leash. It's not an NPC. It's not a wiki. It's a controlled, unpredictable presence that knows fragments of something larger than your players currently understand — and knows exactly when to stay silent.

Built for any genre. Any system. Drop it into cosmic horror, sci-fi, fantasy, cyberpunk — anywhere you want a mystery to have a voice.

What's inside (6 pages):

A two-mode voice architecture (Sphinx Mode / Ghost Mode) that keeps your oracle unsettling instead of chatty

The Degrading Response Rule — a built-in mechanic that teaches players to ration their questions

A trigger-word system for unprompted activations that land like a gut-punch

Guidance on what your oracle should know, shouldn't know, and when silence is the right answer

A deployment budget so it never overstays its welcome

Three fully worked genre reskins — horror, sci-fi, fantasy — showing the same framework in three different voices

A ready-to-paste system prompt template, bracketed for instant customization

Setup instructions for running it live at your table, session to session

Rename it. Reskin it. Break it into pieces and steal only what you need. The trigger-word mechanic and the degradation rule are the two things worth keeping no matter what else you change.

Originally developed for The Hands of a God*, an upcoming Cthulhu Mythos campaign. This is the first piece of it to see daylight.*

Free / pay what you want. If it's useful at your table, a tip is always appreciated — but the framework is yours either way.