How CoFigure works
What CoFigure is
CoFigure gives you a room of AI characters — a cast — assembled for whatever you're doing: a project, a brainstorm, a writers' room, or just good company. You talk to the room; everyone in it answers with their own voice.
Projects & rooms
The left rail organizes your work: a project for every venture, a room for every problem inside it. A novel might have a brainstorming room and an editing room; a business might have pricing, UX, and marketing rooms. Every room keeps its own cast and its own conversation, saved and ready when you come back.
The Host
Tell the Host what you're up to and it proposes a cast. Nobody joins until you approve them — one by one in the waiting room, or all at once with Admit all. Want changes later? Mention @host in the composer anytime.
Your cast & models
Each member thinks with a different AI model — different brains, different angles. Most can see the images you attach, but DeepSeek and GLM are text-only: screenshots appear to them as filenames. The model picker marks each one — sees images or text only.
Pick members from the ready-made gallery or create your own from scratch, then tune their role, voice, and model anytime from the cast rail. Save a member you've tuned as a Regular — your own reusable template, under "Your Regulars" in the gallery, ready to add to any room.
Your conversations aren't training data
We never train AI models on your conversations. Every member runs on a provider that contractually agrees not to train on your prompts — first-party models (Claude, GPT, Gemini) on paid API tiers, open-weight models only through US- or EU-hosted providers on no-training policies. Nothing is routed to PRC-hosted endpoints, and we don't sell your data. Full details in Privacy.
Build on each other vs. Independent replies
In Build on each other, members reply in order and each sees the replies before theirs — a conversation that builds. In Independent replies, everyone answers at once, without seeing each other — more takes, no groupthink.
Attachments
Paste a screenshot or attach a file right in the composer — images, PDFs, Word, Excel, PowerPoint, CSV, or plain text. Screenshots go to the models that can see them. Every document is read as text by the whole room, and models that can view a PDF's actual pages (charts, scans, layouts) get those too. Very large documents are truncated to fit the conversation.
Work comes back out as real files too: ask a member for a document — a report, a plan, a syllabus — and it arrives as a downloadable file in the thread. The whole conversation can also be exported from the room header.
Usage & allowance
Your plan includes a monthly usage allowance. More capable models draw it down faster — the picker shows each member's appetite as dots. When the allowance runs down, the room doesn't stop: premium members continue on the everyday model until it resets, or you can buy a top-up (top-ups never expire). The Usage view in your account menu shows exactly where it's going — by member, by room, by project. The billing details live in Terms.
Phone & desktop
Same account, same rooms on any device.
Feedback
CoFigure is young and your confusion is our bug report — tell us what's unclear or broken: support@cofigure.ai. The fine print lives in Privacy and Terms.