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

# Getting started

> Get access, find your way around the Command Center, and run your first session.

Aperium runs in your web browser — there's nothing to install. This page gets you signed in, oriented on the screen you land on, and through your first real request. The whole trick to a good first session is to start small.

## Get access and sign in

<Steps>
  <Step title="Open your emailed link">
    Access arrives as an **emailed link**. Open it and sign in with your Hillspire login.

    <Warning>
      Aperium does **not** appear as an Okta tile today, so **bookmark the link** — it's your front door back in.
    </Warning>
  </Step>

  <Step title="Follow the onboarding flow">
    The first time you open Aperium, a short guided onboarding walks you through connecting your systems.
  </Step>

  <Step title="Connect Google first">
    Connect **Google Workspace** first — it powers the Command Center, the daily overview you'll land on every time.
  </Step>
</Steps>

<Note>
  Aperium lives at [hillspire.app.aperium.ai](https://hillspire.app.aperium.ai/) — sign in with your Hillspire login and **bookmark it** (it's not an Okta tile today, so the bookmark is your front door). You'll also get the link in your onboarding email.
</Note>

<Note>
  Every system you connect **authenticates and acts as you**, so your existing permissions carry over — you only ever see what you'd already see in that system directly. See [Connected AI](/using/connected-ai) for the why.
</Note>

## Your home base: the Command Center

The **Command Center** is the screen you land on each time you open Aperium. It's a daily overview that brings three things together in one place:

* Your **calendar**
* Your **emails**
* **Suggested replies** Aperium has already drafted for you

The move worth remembering: from any email, hand it straight into a full agent session with **"follow up with agent"** — no copy-paste, no re-explaining the context. That hand-off is what makes the Command Center more than just another inbox.

## Start your first session

A **session** is one continuous conversation thread that keeps its own context. Starting a new one is like opening a fresh notebook page — it doesn't erase anything you did before.

<Info>
  Exact button labels and the interface layout are being finalized against the live product. The interface groups a few regions you'll come to recognize: your past conversations, the box where you type a request, an indicator of which agent you're using, and a signal for when Aperium is doing something versus just answering.
</Info>

### Check which agent you're talking to

Aperium isn't a single assistant — there's a general-purpose agent plus specialized ones. Before asking something specialized, check which agent is active, because a specialist may reach systems or data a general agent can't. If you're not sure which to use, start with the general agent and ask it to point you to the right one.

### Phrase a good first request

For your very first request, pick something that is:

<CardGroup cols={3}>
  <Card title="Concrete" icon="crosshairs">
    A specific ask, not "what can you do?"
  </Card>

  <Card title="Small" icon="minimize">
    Scoped tight, so it's easy to judge.
  </Card>

  <Card title="Already knowable" icon="circle-check">
    Something you can verify yourself.
  </Card>
</CardGroup>

A good pattern: *"Summarize the last 3 messages in \[a channel or doc I already know]"* rather than an open-ended prompt. That way you can tell whether Aperium got it right.

## Talking vs. doing: spotting a tool use

Aperium works in two modes, and it's worth learning to tell them apart:

* **Just answering** — it's composing text from what it already knows. Fast, no visible pause. Treat this like any unverified claim; it's the model's own recall.
* **Using a tool** — it's looking something up, querying a system, or running a search. There's usually a visible signal: a status line, a named action, or a short pause before the final answer lands.

Why it matters: an answer that came from a tool pulling real Hillspire data is checkable against that system. A plain answer is not.

<Info>
  The exact on-screen signal for tool use is being confirmed against the live interface. Watch for a brief status line or named action before the final answer.
</Info>

## First-session tips

<AccordionGroup>
  <Accordion title="Don't lead with something huge or vague">
    A big, open-ended first message makes it hard to tell whether it worked. Start small so the result is easy to check.
  </Accordion>

  <Accordion title="Don't assume it remembers another session">
    Each session keeps its own context. It won't know something from a different session unless you tell it.
  </Accordion>

  <Accordion title="If the first answer looks off, that's normal">
    Rephrase and try again rather than assuming something's broken.
  </Accordion>
</AccordionGroup>

## Share a chat thread

Every session has a **Share** button that shares the whole conversation, the same way you'd share a Google Doc. Two everyday uses:

* **Hand off your work.** Send a colleague the full thread so they can see your analysis and results and pick up from there — no re-explaining.
* **Get help.** If something breaks or a session misbehaves, share the thread into a support channel so whoever's helping can see exactly what happened.

<Info>
  The exact location of the Share button is being confirmed against the live interface.
</Info>

## Next

* Explore who else you can work with in the agent catalog.
* Deepen your prompting technique.
* See the full picture of what connected AI unlocks in [Connected AI](/using/connected-ai), or check the [FAQ](/using/faq).