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

# The Assistant

> A chat copilot beside your editor that builds, explains, and fixes the Space you're in.

The **Assistant** is a chat panel alongside your editor that can build and change the
[Space](/concepts/spaces) you're working in. Describe what you want; it proposes a plan,
shows you each asset before it's written, and creates the files once you approve.

It runs on the **Viorant** managed provider and spends
[credits](/providers/credits) from the same wallet as everything else.

## Where it lives

The Assistant is a collapsible panel on the right-hand side of the Hub. It works on the
**Space you currently have open** — open a Space and it's ready; without one it has
nothing to act on.

Everything it does is scoped to that active Space. It won't read from or write to any
other Space.

## How a build goes

<Steps>
  <Step title="Describe the goal">
    Say what you want in plain language — *"a skill that summarises my unread Slack
    mentions each morning"*. You can also start from the goal box when you create a new
    Space.
  </Step>

  <Step title="Read the plan">
    The Assistant proposes a plan before it changes anything. It's shown as steps you
    can read, and it will ask about anything genuinely under-specified rather than
    guessing.
  </Step>

  <Step title="Approve each piece">
    Each step arrives as a card showing the **actual thing being made** — the prompt
    text, the skill body and the connectors it needs, the model and tier, and how the
    agent gets wired together. You approve them one at a time, or let it build the rest
    without asking.
  </Step>

  <Step title="Watch the files land">
    Approved assets appear in the Space's file tree as they're created. They're ordinary
    files — the same ones you'd have authored by hand. From there you can run and test
    the agent as you would any other.
  </Step>
</Steps>

## Fixing and improving

Once a Space has assets in it, the Assistant can also work on what's already there —
describe what's wrong or what you want changed and it proposes edits the same way, as
reviewable steps.

<Note>
  The **AI prompt builder** in the prompt editor is still there and unchanged. Use the
  Assistant when you're working across a Space — several assets, or wiring them
  together. Use the prompt builder when you want to generate or improve one prompt
  in place.
</Note>

## Connectors and models

If a skill needs a connector, the Assistant brings it in from the connectors already
available to you, or tells you to add one — it can't sign you into a service on your
behalf. See [Connectors](/concepts/connectors) for how authorization works.

For models it defaults to the Viorant managed provider on the **Medium**
[tier](/providers/credits#tiers), and shows you that choice before wiring it in.

## What it costs

Every turn spends credits, including turns that only plan or explain. The
[tier picker](/providers/credits#tiers) in the panel controls how much — **Low** is
cheaper and usually fine while you're still shaping the idea.

If your balance reaches zero the Assistant stops rather than failing quietly. See
[Credits](/providers/credits).

<Card title="Credits & the Viorant provider" icon="coins" href="/providers/credits">
  What spends credits, and what the starting grant covers.
</Card>