> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gobihq.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Personal agent

> Ask anything across everything you've captured.

Your personal agent lives in **AI Chat**, and it knows one thing no
other AI does: your actual life, as you've captured it. Ask it
anything across everything in your core — it answers from what
really happened, not from what it can guess.

<Frame>
  <img src="https://mintcdn.com/gobi/kPhulyAoKO8WjLxE/images/personal-agent-answer.png?fit=max&auto=format&n=kPhulyAoKO8WjLxE&q=85&s=2eb1fc463a4e4201832237cd029c48e9" alt="The personal agent answering from captured conversations" width="402" height="874" data-path="images/personal-agent-answer.png" />
</Frame>

## What to ask it

The best questions are the ones your memory fumbles:

* "What did Dana say about pricing?"
* "What did I promise this week, to anyone?"
* "When did I last talk to Marcus — and what about?"
* "What's still open from Tuesday's meeting?"

Answers come with their receipts: the agent draws on your
captures, so what it tells you traces back to conversations you
actually had. Type or talk — it takes both.

## It does things, too

Ask for more than recall and it works in front of you — you'll see
it thinking, then answering, and you can reply mid-stream to steer
it:

* "Brief me on everything with Cintra before my 2pm."
* "Draft a follow-up to yesterday's meeting covering what we
  agreed."
* "Pull the action items from this week into one list."

The more precise the ask, the better the work — it's an agent, not
a search box.

## Why it gets smarter

The agent's ceiling is your core. Every capture adds context it
can draw on: more conversations, sharper answers; more history,
better briefs. A fresh core gives thin answers — that's not a flaw,
it's day one. Capture for a week and ask again; the difference is
the product.

<Tip>
  The fastest way to feel it: after your next three meetings, ask
  "what did I commit to this week?" — then imagine answering that
  from memory.
</Tip>

## Yours alone

The personal agent reads your core and nothing else — not your
teammates' captures, not what hasn't been shared with you. And the
conversation flows one way: chatting with your agent never shares
anything outward, and
[nothing trains on your data](/privacy/models-and-training) unless
you switch that on.

Want the same context in the agents you already run — Claude Code,
Codex, Cursor? That's
[Connect your agent](/getting-started/connect-your-agent): your
core, piped into your stack.