What we do, and deliberately do not do, with your writing
Writers are right to be suspicious of tools with AI in them. So here is the whole deal, in plain language, with nothing buried in a terms document.
We believe AI belongs in writing as an aid, not an author. It is genuinely useful for the work around the words: filing your lore, cross-linking your codex, catching continuity slips, helping you get unstuck. We do not recommend using it to write your prose outright. The voice that makes a story worth reading is yours, and no model has one.
We know it is a little hypocritical to say that while shipping an AI drafting feature. But our job is to provide every tool a writer could want and leave the choice of which ones to use where it belongs, with you. If AI drafts are part of your process, treat them as scaffolding: argue with them, rework them, overwrite them. The words you keep should be yours.
That choice is built into the app itself. Every feature you do not want is a toggle away: switch it off and it disappears from your workspace entirely, not grayed out, gone. Switch it back on whenever you like. You are in full control of which tools you ever see.
LoreLynx does not train AI models on your writing, sell it, or license it for training or advertising. Service providers process information needed to run LoreLynx. If you use a connected cloud AI provider, its handling of your request depends on its terms and your account settings, including retention, review, and training.
Every feature of LoreLynx works with AI disabled: the story planner, the prose editor, the codex, maps, timelines, exports, all of it. If you never connect a provider, nothing you write is ever sent to any AI, anywhere. There is no hidden analysis, no background processing, no fine print.
Connect LM Studio or Ollama running on your computer and supported local AI requests go directly to that model. Your saved projects still sync to LoreLynx in the cloud. An endpoint on another machine is not necessarily local.
You obtain your own provider account and key separately. Keys are stored encrypted in your browser and are not persistently stored on our servers. Cloud requests transmit your key and relevant content through our server to your selected provider. We do not persistently store the relay request body; work you save remains in your project.
Any passage of prose and any codex entry can be excluded from AI context. Excluded material is stripped out before anything is sent to a model, even when AI features are on. Spoilers, private notes, and unfinished ideas stay between you and the page. You can also see exactly what was sent with any generation via the context breakdown.
Export your manuscript as PDF, DOCX, EPUB, or Markdown, and project data as JSON you can back up or re-import. JSON and Markdown exports are on every plan; DOCX, EPUB, and PDF exports are part of Pro.
Besides cloud sync, LoreLynx automatically keeps recent snapshots of your project on your own device. You can download or restore them from Settings without needing our servers at all.
Share links show readers your prose and chapter structure, nothing more. Your planning notes, summaries, beats, and drafts are never exposed through a share link, and you can revoke any link whenever you like. On the book feed, a codex entry reaches readers only when you publish it yourself, one entry at a time, to your book's World page; your notes never do. Anyone holding a valid share link can use it. Revocation cannot recall copies a reader saved.
Account deletion removes your account-owned work from active systems; if a step fails, your account is left unchanged and you can try again. Restricted database backups rotate over seven days. Limited records and contributions to another person's project may remain as described in our Privacy Policy. Deletion cannot recall copies already saved by others.
The formal versions of these promises live in the Privacy Policy and Terms. If anything here ever seems out of step with how the app behaves, tell us. That would be a bug, and we treat it like one.