Current context in reopened chats
Source changes reach the context of existing conversations. Check the cited source and original content when an answer depends on a recently updated document.
Second Brain
Preserve what matters and find it when you need it. Create independent personal or business Brains, choose what to import and connect only the archives relevant to your work.
T-CHAT 4.6.5
Source changes reach the context of existing conversations. Check the cited source and original content when an answer depends on a recently updated document.
Import .md files too. Email, calendars and content from supported Office accounts can be selected through preview and import; connecting an account does not automatically synchronise all its content.
Request an update during a chat or review an AI proposal. Adding information to Second Brain requires confirmation and remains distinct from generated conversation content.
The 4.6 series includes profiles and visual analysis, subject to the supported formats and limits. Connected account sources are selected and approved before import.
Memory under your control
A personal Brain for your writing, a business Brain for procedures and others for projects. Access and sources stay separate: one archive’s contents do not automatically become available to another.
Search and answers
Search text and, if you enable a local embedding model, retrieve by meaning as well. Answers draw content from the selected archive and provide references for review.
Personal profiles
Add your history, values, preferences, decision criteria, style and goals. Choose whether answers use them and keep revisions. Separate Brains can have separate profiles, preserving what you have already built.
Backup and recovery
Export encrypted .tbrain copies and optionally configure scheduled backups and retention. Restoring creates a new Brain rather than overwriting the one you are working on.
Yes. Import is explicit and retains sources in the selected archive. Storing a file does not mean every format can be read or indexed: supported formats and limits are checked during ingestion.
It can represent knowledge, declared memories and preferences through sources and a profile. It is not a complete replica of a person, does not train the model and can make mistakes.
Visual reading analyses supported single images and up to 20 PDF pages per request, within 100 MB per file and 32 megapixels per image. HEIC, animations and audio/video transcription are not included. A visual description does not establish identity or memories.
.tbrain backups are encrypted. This is not the same as encrypting the entire active archive at rest: disk protection, access to the Mac and key custody require appropriate configuration.
Use cases
Choose an activity, prepare a non-sensitive data sample and let’s define the expected outcome, hardware and checks together.