> ## 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.

# Connect your agent

> One paste, and the agent you already use knows your real life.

Your agent knows everything you've typed and nothing you've lived.
This page fixes that. Connecting takes one paste: Gobi gives you a
short instruction, you hand it to your agent, and your agent does
the rest — installs the Gobi CLI, authenticates, and confirms the
connection. From then on, your personal core is context it can
draw on.

<Info>
  Works with any agent that can run commands: Claude Code, Codex,
  Cursor, and most agentic setups. If yours can use a terminal, it
  can connect.
</Info>

## Connect in three steps

<Steps>
  <Step title="Open the connect sheet">
    In the Gobi app, choose your agent when asked **"Which agent do
    you work with?"** — during onboarding, or any time from
    **Account settings → Connect with agents**. Gobi generates your
    personal connect instruction.
  </Step>

  <Step title="Paste it into your agent">
    Copy the instruction and paste it to your agent as a message.
    It looks like this:

    ```text theme={null}
        Install the Gobi CLI from github.com/gobi-ai/gobi-cli,
        authenticate with token gbi_xxxxxxxxxxxx, and confirm
        the connection.
    ```

    Your agent installs the CLI, authenticates with your token, and
    confirms. That's the whole setup.

    <Warning>
      The token is single-use and expires in 24 hours. If it
      expires before you connect, generate a fresh one from the
      same sheet — and treat it like a password: it links an agent
      to *your* core.
    </Warning>
  </Step>

  <Step title="Ask something only your life can answer">
    Try it immediately:

    ```text theme={null}
        What did I capture yesterday?
    ```

    ```text theme={null}
        What did I promise in my last meeting?
    ```

    If your agent answers from your real day, you're connected.
    From now on, every capture makes it smarter.
  </Step>
</Steps>

## What your agent can access

Your agent reads from your personal core: your captures, their
Summaries, attendees, and your notes. It sees what you see in
Home — nothing more. It can't see other people's cores, can't see
what teammates haven't shared with you, and connecting an agent
never shares anything *outward*: the pipe flows from your core to
your agent, not the other way.

The full picture lives in
[What agents can and can't see](/privacy/what-agents-can-see).

## Managing connections

Connect as many agents as you like — each connection uses its own
single-use token, generated from **Account settings → Connect with
agents**. Tokens expire 24 hours after they're issued and can't be
reused, so an old instruction can never open a new connection.

## Troubleshooting

**The token expired.** Tokens are single-use and live 24 hours.
Generate a fresh instruction from the app and paste again.

**My agent says it can't install the CLI.** It needs terminal
access. Agents that only chat — without the ability to run
commands — can't connect today.

**It connected, but answers seem empty.** A young core is a quiet
core. Capture a few real conversations and ask again; the answers
grow with what you've lived.