> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a2v2.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Build, train, and embed your first AI agent in about 5 minutes.

This guide takes you from an empty account to a working, embeddable AI agent.
You'll create an agent, give it knowledge, test it, and grab the embed snippet.

<Note>
  **Before you begin** — sign in at [app.a2v2.ai](https://app.a2v2.ai). New here?
  Sign up with your email (you'll get a one-time code) or continue with Google.
  See [Signing in](/concepts#accounts) for details.
</Note>

## 1. Create an agent

From the **Agents** dashboard, select **New Agent**. In the **Add Agent** dialog,
choose an **agent type** (pick **General Agent** for most use cases), give it a
name, and select **Create**. You'll land on the agent's **Playground**.

<Tip>
  Name agents by their job — "Support Bot", "Onboarding Assistant" — so they're easy
  to find once you have several.
</Tip>

[Full guide: Create an agent →](/building/create-agent)

## 2. Add your first source

Open **Knowledge Base** from the agent's sidebar (under **Expertise**), select
**New Knowledge Item**, and pick a source type:

<Tabs>
  <Tab title="Files">
    Open the **Files** tab and choose a PDF, DOCX, or TXT. The file is uploaded
    and queued for processing.
  </Tab>

  <Tab title="Links">
    Open the **Links** tab, enter a website URL, and A2V2.ai indexes the page
    content.
  </Tab>

  <Tab title="Q&A">
    Open the **Q\&A** tab and add a question and its answer directly — great for
    policies and FAQs.
  </Tab>
</Tabs>

Each source shows a status as it's indexed. Wait for it to reach **Completed**
before testing.

<Info>
  Processing happens in the background — you can keep working while a large file
  indexes. See [how processing works](/knowledge-base/overview#how-processing-works).
</Info>

[Full guide: Knowledge Base →](/knowledge-base/overview)

## 3. Test your agent

Go to the **Sandbox** and ask a question your source can answer. The agent
streams its reply and shows the **sources** it used, so you can verify the answer
came from your content.

<Tip>
  If the answer is thin or wrong, your source may still be processing, or the
  content may not cover that question. Add more sources or sharpen your
  [instructions](/building/instructions).
</Tip>

## 4. Customize the look (optional)

Open **Agent UI → Appearance** to set the agent's colors, avatar, and the chat
bubble position so the widget matches your brand.

[Full guide: Appearance →](/customize/appearance)

## 5. Embed it on your site

Open **Channels & API → Embed/Install**. You have three options:

<CardGroup cols={3}>
  <Card title="Share Link" icon="link">
    Send a hosted link to your agent — nothing to install.
  </Card>

  <Card title="Chat bubble" icon="comment">
    A floating button in the corner of your site. Add the widget script and
    config to your page's HTML.
  </Card>

  <Card title="Inline iframe" icon="window-maximize">
    Embed the chat directly inside a page. Copy the `<iframe>` snippet where you
    want it to appear.
  </Card>
</CardGroup>

Add the snippet to your site and reload — your agent is live.

[Full guide: Embed & Install →](/embed/install)

## Next steps

<CardGroup cols={2}>
  <Card title="Write better instructions" icon="pen" href="/building/instructions">
    Give your agent a persona, tone, and rules to follow.
  </Card>

  <Card title="Choose the right model" icon="microchip" href="/building/model-settings">
    Balance answer quality, speed, and credit cost.
  </Card>

  <Card title="Add more knowledge" icon="layer-group" href="/knowledge-base/overview">
    Files, websites, YouTube, and Q\&A — all in one place.
  </Card>

  <Card title="Compare configs" icon="code-compare" href="/testing/playground">
    A/B test instructions and models in the Playground.
  </Card>
</CardGroup>