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

# HelpAlive Agent: in-app AI assistant for SaaS

> The HelpAlive Agent is a context-aware, in-app AI assistant trained on your docs that answers user questions in real time across your SaaS product.

The **Agent** is HelpAlive's in-app AI assistant. It lives inside your product, knows the user's current page, role, and recent activity, and answers their questions grounded in your documentation — in real time.

You configure the Agent from the **Agent** section of your dashboard. There are five tabs:

| Tab                 | What you do here                                                                                 |
| ------------------- | ------------------------------------------------------------------------------------------------ |
| **Train**           | Add your knowledge sources — docs URL crawls, PDFs, Word documents, or pasted text.              |
| **Personalization** | Set the Agent's tone, length, and custom rules. Add specific answers for special-case questions. |
| **Test**            | Run a batch of test questions against your draft changes before you publish.                     |
| **Appearance**      | Match the Agent to your brand — colors, position, welcome message, layout.                       |
| **Configuration**   | Visibility, rate limits, and danger-zone reset controls.                                         |

## What your team gets

<CardGroup cols={2}>
  <Card title="A real-time helpdesk inside your product" icon="message-circle">
    Users ask questions in plain language, in the moment of confusion. The Agent answers grounded in your docs and the user's current situation — not a generic FAQ.
  </Card>

  <Card title="Lower support volume on routine questions" icon="trending-down">
    Anything covered by your knowledge base resolves before it becomes a ticket. What does escalate arrives with full context attached.
  </Card>

  <Card title="A signal of what's missing" icon="inbox">
    Every question the Agent couldn't answer well lands in the [Topics inbox](/studio/topics) — your team's working backlog of "what to write next."
  </Card>

  <Card title="A path to running [Guides](/studio/guides)" icon="map">
    When a user's question is "how do I do X?", the Agent identifies the goal and (soon) launches the matching Guide instead of just describing the steps.
  </Card>
</CardGroup>

## What makes the Agent context-aware

A generic AI chatbot reads the message and replies. The HelpAlive Agent does more: every answer is grounded in **four kinds of context**, in real time.

<CardGroup cols={2}>
  <Card title="Page context" icon="window">
    The exact page the user is on, the title, and what's visible on screen.
  </Card>

  <Card title="User identity" icon="user">
    Who the user is, their role, plan, and which workspace they belong to — straight from your `identify()` call.
  </Card>

  <Card title="Recent activity" icon="clock">
    What they clicked, where they got stuck, and which errors they saw in the last few minutes.
  </Card>

  <Card title="Your knowledge base" icon="book-open">
    Your docs, help articles, and custom answers — searched for relevance before every reply.
  </Card>
</CardGroup>

The result: instead of opening with "How can I help?", the Agent says something like *"I see you were trying to add a teammate from the Billing page — want me to walk you through it?"*

## What your end users see

When you turn the Agent on, your users see a launcher button in the corner of your product. When they open it, they get:

* A **personalized greeting** based on who they are and what page they're on. ("Hey Priya — anything I can help with on the Workflows page?")
* **Three suggested questions** for the current page, picked from what users in similar situations actually ask.
* A **real-time conversation** with thumbs-up / thumbs-down feedback, so you can see which answers worked.
* A **start-fresh** option any time the conversation drifts off topic.

End users don't install anything. The Agent appears automatically once HelpAlive is integrated.

## What you control

Every part of the Agent's behavior is configurable from the dashboard, with a draft / publish workflow so you can preview changes before they reach end users.

* **Knowledge sources** — what the Agent reads to answer questions. Crawled docs, uploaded files, or pasted text. See [Train](/agent/train).
* **Tone and rules** — formal vs casual, concise vs thorough, custom instructions like "always link to a doc" or "never recommend competitor products." Set in **Agent → Personalization**.
* **Custom answers** — for things that aren't doc-shaped. Internal policies, plan-specific responses, time-sensitive notices. Configured in **Agent → Personalization**.
* **Appearance** — color, position, welcome message, color scheme. See [Appearance](/agent/appearance).
* **Visibility** — turn the Agent off entirely, scope it to specific workspaces, or hide it for users who haven't consented to chat tracking.

## At a glance

|                         |                                                                                      |
| ----------------------- | ------------------------------------------------------------------------------------ |
| **Where it runs**       | Inside your product, in a chat panel users can open from any page                    |
| **What it looks like**  | A clean chat widget that picks up your brand colors and won't clash with your styles |
| **How fast**            | Replies appear in real time, word by word                                            |
| **What it learns from** | Your docs site, uploaded files (PDF, Word, Markdown), or pasted text                 |
| **Setup time**          | Around 10 minutes                                                                    |
| **Workflow**            | Edit → preview on the **Test** tab → publish when ready                              |

## Privacy

Conversations live in their own secure store and **never** mix with your behavioral analytics. Personal data in messages is automatically redacted before storage. See [Privacy](/privacy/overview) for the full picture.

## Next steps

<CardGroup cols={2}>
  <Card title="Train on your docs" icon="book-open" href="/agent/train">
    Add your first knowledge source.
  </Card>

  <Card title="Personalization" icon="message-square" href="/agent/personalization">
    Set tone, write custom answers, and tune the guidance templates.
  </Card>

  <Card title="Test" icon="flask-conical" href="/agent/test">
    Preview your draft before users see it.
  </Card>

  <Card title="Customize appearance" icon="palette" href="/agent/appearance">
    Match the Agent to your brand.
  </Card>

  <Card title="Configuration" icon="sliders-horizontal" href="/agent/configuration">
    Turn the Agent on or off, set rate limits, reset training.
  </Card>

  <Card title="Publishing changes" icon="git-branch" href="/agent/publishing">
    Stage your changes, preview, then publish.
  </Card>
</CardGroup>