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

# Get Started

> Connect WhatsApp, teach your agent about your restaurant, and switch it on safely in shadow mode — all in about ten minutes.

Getting your AI agent live takes three steps: **connect WhatsApp**, **teach it about your restaurant**, then **switch it on** — starting safely in shadow mode so you approve every reply at first.

<Steps>
  <Step title="Connect WhatsApp">
    The agent answers customers on WhatsApp, so you need a connected WhatsApp number first.

    If you haven't connected one yet, open **Settings → AI Agent** and you'll see an invitation to connect it. Click **Connect WhatsApp** to begin.

    <Frame caption="Before WhatsApp is connected, the AI Agent page invites you to set it up.">
      <img src="https://mintcdn.com/eatnow/GZg_M_tiiCZcXUbZ/images/screenshots/agent-empty-state.png?fit=max&auto=format&n=GZg_M_tiiCZcXUbZ&q=85&s=7e5b53d220e98f14fdc870516d98fc72" alt="The AI Agent empty state inviting you to connect WhatsApp" width="1534" height="1152" data-path="images/screenshots/agent-empty-state.png" />
    </Frame>

    <Note>
      Already messaging customers on a WhatsApp Business number? You can keep it. Follow the [Connect WhatsApp Business](/whatsapp/onboarding) guide — we recommend the **Coexistence** option so you keep your existing number.
    </Note>
  </Step>

  <Step title="Teach your agent" id="teach-your-agent">
    EatNow already knows your opening hours, party-size limits, deposit policy and contact details — the agent uses them automatically. What's left are the things only you can answer: parking, dress code, terrace, dietary options, and so on.

    From the AI Agent page, click **Run onboarding**. A short guided questionnaire walks you through \~16 common topics, one question at a time. Answer what's relevant, **skip** anything that doesn't apply, and add your own questions at the end.

    <Frame caption="The onboarding wizard — one question per screen, everything is skippable.">
      <img src="https://mintcdn.com/eatnow/GZg_M_tiiCZcXUbZ/images/screenshots/agent-onboarding.png?fit=max&auto=format&n=GZg_M_tiiCZcXUbZ&q=85&s=7053f4dd4f3ae42aa0c16bdb649d29e5" alt="The agent onboarding wizard asking about wheelchair accessibility" width="2000" height="996" data-path="images/screenshots/agent-onboarding.png" />
    </Frame>

    It takes about two minutes. Everything you enter lands in your [knowledge base](/ai-agent/knowledge-base), where you can edit or add to it any time.
  </Step>

  <Step title="Switch it on — in shadow mode">
    Open **Settings → AI Agent** and turn on **Agent enabled**.

    Leave **Shadow mode** on (it's on by default). In shadow mode the agent doesn't message customers directly — it *drafts* a reply and waits for you to approve, edit, or reject it from your inbox. It's the safest way to start: you see exactly how the agent behaves on real conversations before letting it reply on its own.

    <Frame caption="In shadow mode, each agent reply is a draft you approve, edit, or reject.">
      <img src="https://mintcdn.com/eatnow/GZg_M_tiiCZcXUbZ/images/screenshots/agent-draft-review.png?fit=max&auto=format&n=GZg_M_tiiCZcXUbZ&q=85&s=084c95e4f1eaef5a627fce23ee23624c" alt="A shadow-mode draft reply awaiting approval in the inbox" width="2000" height="1063" data-path="images/screenshots/agent-draft-review.png" />
    </Frame>

    <Tip>
      Stay in shadow mode for the first couple of weeks. Once you trust the replies, turn shadow mode **off** and the agent will send directly — you can still take over any conversation at any moment.
    </Tip>
  </Step>
</Steps>

## What happens next

* Customers messaging your WhatsApp number get **instant** replies (drafts for you to approve while in shadow mode).
* You review everything from your **inbox** — see [Reviewing conversations](/ai-agent/reviewing-conversations).
* Anything sensitive is routed to your team automatically — see [Safety & controls](/ai-agent/safety-and-controls).

## Fine-tune before going live

You don't have to configure everything up front, but these are worth a look:

<CardGroup cols={2}>
  <Card title="Give it a voice" icon="pencil" href="/ai-agent/configuration#persona">
    Set the agent's name, tone, and signature.
  </Card>

  <Card title="Choose what it can do" icon="toggle-right" href="/ai-agent/configuration#what-the-agent-can-do">
    Turn booking, edits, and cancellations on or off.
  </Card>

  <Card title="Set its hours" icon="clock" href="/ai-agent/configuration#activation">
    Let it reply 24/7 or only during a schedule.
  </Card>

  <Card title="Test it safely" icon="flask" href="/ai-agent/reviewing-conversations#test-the-agent">
    Send fake messages and watch how it responds.
  </Card>
</CardGroup>