AI writing that can work with your files and context
ChatOTL is a private AI writing assistant built on DeepSeek. It drafts, rewrites, edits and brainstorms from documents you attach, retrieves passages from knowledge bases you have already built, applies presets that hold your standing instructions, and branches alternate drafts so you can compare versions instead of overwriting one.
Most AI writing tools give you a blank prompt box and no memory. Every draft starts from nothing, so you re-paste the brief, the style rules and the source material every time.
ChatOTL is built the other way round: your files, your reusable knowledge and your saved instructions are part of the workspace, so a draft starts from your context rather than the model's training data.
What the assistant does
- Drafting — full pieces, sections, emails, outlines and alternative openings.
- Rewriting — shorter, plainer, for a different reader, in a different register.
- Editing — line edits, consistency passes and structural critique of a draft you already wrote.
- Brainstorming — angles, headlines, counter-arguments and objections to your own thesis.
Writes from your documents
Attach up to 5 files per message at 15 MB each — briefs, transcripts, specs, Markdown drafts, CSV data. Text, Markdown, CSV, JSON and source files are extracted inline; PDFs and images go through a separate vision pipeline that transcribes them, and that transcription is what DeepSeek reads.
Reusable context, not repeated prompts
- Knowledge bases: upload style guides, product docs or past work once and have relevant passages retrieved into later drafts.
- Presets: save the instructions you always give — banned phrases, house style, formatting — and apply them to any new conversation.
- Persistent threads: one conversation per piece, scoped to your account, so context accumulates instead of resetting.
- Keyword search across your own message history to find the version you wrote last month.
Branch alternate drafts
Fork any reply into a new thread to develop a second version of the same passage. Both survive, so you choose between drafts rather than talking one draft into a corner. Model choice is yours mid-draft: deepseek-v4-flash for speed, deepseek-reasoner when the structure needs argument.
Private by construction
Row-level security scopes every conversation, message, setting and uploaded file to the account that created it. You can supply your own DeepSeek key or register a compatible endpoint of your own.
Limitations
- No web browsing, so it cannot check a current fact, price or source for you.
- Facts, quotes and figures need your verification before anything goes out.
- It approximates a voice from samples you supply; it does not know your voice by default.
Frequently asked questions
Can ChatOTL help revise existing writing?
Yes. Paste or attach the draft and ask for a line edit, a structural critique or a rewrite at a specific length. Revision is the strongest use of the assistant, because it works on text you already own.
Can I upload reference documents for it to write from?
Up to 5 files per message, 15 MB each: PDFs, images, plain text, Markdown, CSV, JSON and source code. Text-based files are extracted inline; PDFs and images are transcribed by a vision pipeline first, and DeepSeek reads that transcription.
Can I save writing instructions so I stop retyping them?
Yes — save them as a system-prompt preset and apply it to any conversation. For larger reference material, a knowledge base is the better home: upload once, retrieve into later drafts.
Which model writes the text?
DeepSeek. You can switch between deepseek-v4-flash, deepseek-chat and deepseek-reasoner mid-conversation. Gemini is used separately for embeddings and for transcribing images and scanned PDFs — DeepSeek never sees an image directly.
Can I keep two versions of a draft?
Branch the reply. Forking starts a new thread from that point, so the alternate version develops without destroying the original.
Start using ChatOTL for writing
Create a free account, attach your source material and draft from your own context.
Start using ChatOTL