FAQ

Straight answers.

Being honest about the AI is one of our four founding rules. If your question isn't here, ask — skeptics welcome.

The AI, plainly

What does the AI actually do?

It writes narration text, live. Three reactive handlers narrate combat, scene activations, and Party Lore journals as they happen; nine chat commands generate NPCs, encounters, loot, rules answers, recaps, and session prep when you ask. Generation runs on a hosted large language model (Anthropic's Claude) — nothing runs on your machine.

Will it ever generate art?

No. Not tokens, not portraits, not maps — the engine has no image tools and declines art requests by design. Art comes from human artists; our directory sends GMs to them. The full reasoning is on the About page.

Does it act on my world without me?

It narrates in chat when game events you cause happen, and it can create documents (an NPC you asked for, a recap journal). It cannot delete anything — the capability doesn't exist — and reactive lore only ever reads journals you put in the Party Lore folder.

You generate settings now — isn't that a “content mill”?

No — and the distinction matters to us. The Setting Forge is GM-invoked prep tooling: you seed it, approve every section, and the result lands in your own world as your own campaign material. We never sell or ship content packs. Inspired mode extracts only the structure and tone of what you name — it refuses to reproduce copyrighted worlds, characters, or names, the same way the engine refuses art. Public-domain and your-own-notes modes build faithfully from material you attest to. And as always: the words are AI, the art never is.

Can my players drive it?

Commands are the GM's. The module connects from the GM's client, and generated GM material (continuity reports, condition reminders) is whispered to the GM alone.

Privacy

Can it read my GM whispers?

No. Whispers are filtered out on your server before anything syncs, and filtered again on the live chat feed. Secret condition notes and private asides never reach the engine, so they can't leak into player-facing text.

What does the service store?

A short-lived working copy of your world (actors, scenes, journals, recent table chat) that's rebuilt every time the module connects — your Foundry server stays the source of truth. Your generation history is kept so your dashboard can show it, along with what each generation drew on.

What should I keep out of the Party Lore folder?

Anything players shouldn't see summarized. That folder is the one place reactive lore reads and posts from — prep notes anywhere else are never summarized to the table.

Game systems

Which systems are supported?

Adapters ship for D&D 5e, Pathfinder 2e, Blades in the Dark, Call of Cthulhu 7e, and Cyberpunk RED — stat blocks, difficulty scaling, and combat beats speak each system's language. Call of Cthulhu even gets sanity-aware whispers to the Keeper.

What if my system isn't on that list?

The module itself is system-agnostic, so narration and commands still work in a generic voice — unknown systems degrade gracefully rather than erroring. Adapter coverage grows from what beta tables actually run, so tell us your system when you join.

Where do rules answers come from?

An ingested rules corpus, looked up fresh on every !rules question and quoted back — not the model's memory of a rulebook.

Cost

What counts as a generation?

Each piece of text the engine writes: one round's combat narration, one scene description, one NPC, one recap. Chat commands, reactive narration, and the web dashboard all draw from the same monthly pool — there are no separate meters.

What happens when I hit the cap?

Reactive narration pauses with a single notice in chat, and commands tell you you're capped. Your world, your data, and the module keep working, and the count resets monthly.

Do I need my own AI API key?

Never. The engine is hosted — your gmg_ key is just your GM Grimoire account key. Flat plans, plain generation counts, no token math.

What does the beta cost?

Nothing — no card required. Free (20 generations/month) and Pro ($9/month, 300) switch on after the beta proves them. Plans and details live on the Pricing page.

Deeper dives: Features ·Pricing · Why no AI art ·The Foundry module