AI chat with the controls a daily driver needs
AI chat is a conversational interface to a large language model: you type a message, the model answers with the thread as context. ChatOTL adds streaming output, mid-conversation model switching, branching and file attachments, inside a workspace scoped to your account.
The basic loop of AI chat is identical everywhere. What separates a demo from a tool is the control you have over the loop.
ChatOTL is built around four of those controls: stream, switch, branch and attach.
Stream
Tokens render as they arrive so you can judge a direction early. Stop mid-answer when it's wrong, regenerate when it's close.
Switch
Pick a different model for the next turn without losing context. Models from OpenAI, Anthropic and Google are supported, plus any OpenAI-compatible endpoint you configure.
Branch
Fork any reply into a new thread. The original stays where it was, and the branch inherits the conversation up to that point.
Attach
Add up to five files per message. Text, Markdown, CSV, JSON and source files are extracted into the prompt; PDFs and images go to the model through short-lived signed URLs.
Frequently asked questions
How does AI chat actually work?
Your message plus the earlier turns are sent to a language model, which predicts a reply token by token. The conversation history is what gives the model context, which is why long threads behave differently from fresh ones.
Is there a free way to try it?
Create an account and start a thread — no provider key is required to send your first messages.
Can I use several models in one conversation?
Yes. The model applies per turn, so you can draft with one and review with another.
Are my chats visible to anyone else?
No. Every thread is scoped to the account that created it by database-level access rules.
Open a thread
Streaming replies, model switching and branching are available from the first message.
Start chatting