How the memory layer works
Characters, locations, continuity — surfaced when you ask, hidden when you don't. Never used as training data.
Arqo's AI is a memory layer. It does not generate dialogue. It does not draft scenes. It remembers what you wrote and answers questions about it.
What it remembers
- Characters — every appearance, every line, where they break
- Locations — every scene there, INT/EXT, time of day
- Threads — A-plot, B-plot, C-plot, where each was last touched
- Continuity — "where did I last mention the locket?" → page 31, scene 17
Per-script. Opt-in per surface.
You can disable the memory layer at the script level (Settings → Memory) or globally (Account → AI). The default is on; nothing about it is hidden.
Never trained on
Vow #2 of the manifesto: scripts never enter training data. Not by Arqo, not by any provider we route through. The transparency log on your account shows every AI call, what was sent (summarized), what came back. Exportable.
BYOK on Pro+
Already paying Anthropic, OpenAI, or Google? Plug your own key in (Account → AI → Provider keys). You keep the spend, the billing, the relationship.