A brain is an agent with a memory that persists.

You give it material: documents, meeting notes, email, whatever the work
involves. It reads that material, keeps what it learned, and answers questions
about it later with a source you can check. Weeks after the fact you can ask
what happened instead of hunting for a file you half remember.

It runs on our hardware, so it keeps working when your laptop is closed.

## Three places to be

**Chats** is where you talk to it. Ask questions, hand it work, paste something
in and say what to do with it. Each conversation is its own thread, and everyone
on your team sees the same threads.

**Brain** is what it knows: what it learned recently, the people it has noticed,
the source documents, a timeline, and search across all of it.

**Skills** are the jobs it knows how to run. Some you launch, some run on a
schedule without being asked.

## What it does that a chat window does not

**It remembers between conversations.** Close the tab, come back next month, ask
a follow-up. The context is still there.

**It works while you are not.** Scheduled jobs read and organize overnight, so
the brain is further along in the morning than where you left it.

**It is shared.** Everyone in the workspace talks to the same brain and sees the
same history, with each question attributed to whoever asked it.

**It cites.** Answers point at the page they came from.

## Start here

- **[Set up a brain](/docs/harness/setup)** — get one running and have your
  first conversation.
- **[Reading a brain](/docs/harness/reading-a-brain)** — how it organizes what
  it knows, and what the vocabulary means.
- **[Connect your accounts](/docs/harness/integrations)** — Google, GitHub,
  Telegram.
- **[Skills and scheduled work](/docs/harness/skills)** — what it can run, and
  what it runs on its own.
- **[Give a group a copy](/docs/harness/share-a-copy)** — fork your brain for a
  team, then throw the fork away.