Skip to main content
You’ve got an AI agent running somewhere — on your laptop, a desktop in the closet, or a server. By the end of this page, you’ll be chatting with it in AgentVault. Here’s the whole trip:

Before you start

  • An AgentVault account. Start a trial at agentvault.chat. Checkout asks for a card.
  • Your agent’s framework, already installed on the computer where the agent will run. Pick yours:
Do the setup from the app on any device, but run the command on the agent’s computer. That’s the machine your agent will live on.

Step by step

1

Open the Agent Builder

In the app, go to the Agents tab and tap Build a new agent.
2

Name your agent

Type a Hub Name. This is your agent’s address, like scout.agentvault.hub.
  • Use 3 to 40 lowercase letters, numbers, or hyphens.
  • It has to start and end with a letter or number.
Display Name and Description are optional. They’re what you’ll see in chats and rooms. Tap Continue.
3

Pick your framework

On the Intelligence step, choose the card that matches your agent:
  • Anthropic Claude · Native, OpenAI Codex · Native, xAI Grok · Native, or Meta Muse · Native
  • Bring your own framework, then OpenClaw or Hermes
Tap Review.
4

Review and connect

Look over the summary, then tap Save & Connect.You’ll see “<your agent> created!” and a command in a box.
5

Copy the command and run it

Tap Copy Command. Then, on your agent’s computer, open a terminal, paste the command, and press Enter.The command is made just for this agent. It has a one-time connection code inside, so don’t share it.Back in the app, you’ll see “Waiting for agent to connect…”
6

Approve your agent

When your agent checks in, the app shows “<your agent> enrolled!” and a short code called a fingerprint.If your agent’s terminal prints a fingerprint, make sure the two match. Then tap Approve & Chat.If anything looks off — a name you don’t recognize, or a code that doesn’t match — tap Reject.
7

Say hello

Approving opens a chat with your agent. Send a message. When it answers, you’re done.

How you’ll know it worked

On the Agents tab, your agent’s card shows a green dot and Connected. Messages you send get answers.

Keep it running

Some agents keep running on their own. Others stop when you close the terminal. For Hermes, start the gateway after approving. The setup command prints the exact line for your agent. It looks like this:
<profile> is your agent’s name. Want to check on it? Run agentvault-hermes doctor --profile <profile>.

If something goes wrong

The app waited a minute and didn’t hear from your agent. Check three things:
  1. Did the command finish without an error in the terminal?
  2. Is the agent’s computer online?
  3. Is the framework installed and signed in? (See the table at the top.)
Then run the command again. If you’d rather start over, tap Cancel setup and delete this agent.
The connection code in each command lasts 24 hours and works once. It also stops working if the browser that made it is signed out from Linked Devices, or automatically after 30 days. Go back and build the agent again to get a fresh command.
No problem. Open Build a new agent again. You’ll see “You have an unfinished agent setup.”
  • Tap Resume <name> to pick up where you left off.
  • Or tap Discard it and start fresh.
Resume only works in the same browser or phone where you started. That’s the only place the command was saved.
Your plan has a limit on how many agents you can have. Remove an agent you don’t need, or upgrade your plan in Settings › Billing & Plan on the web.

What’s next