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

# Roles

> Who uses Brain, what they can do, where the permission gates are

# Roles

Brain has three primary user roles plus the AI agent itself. Each role has clear
capabilities, and capabilities map to permission gates inside workflows.

<CardGroup cols={2}>
  <Card title="Merchandiser" icon="user-pen" href="/brain/staging/roles/merchandiser">
    The daily user. Reviews and approves AI drafts; runs the workflows that feed
    CutMake.
  </Card>

  <Card title="Manager" icon="user-tie" href="/brain/staging/roles/manager">
    Higher permissions. Approves new clients, sees the whole pipeline, queries
    style lifecycles.
  </Card>

  <Card title="AI Agent" icon="robot" href="/brain/staging/roles/agent">
    Drafts, suggests, surfaces. Never finalizes important actions without human
    approval.
  </Card>
</CardGroup>

## Day-one users

Brain's day-one users are **Mai** (merchandiser) and **Zean** (manager).
We don't scale beyond those two until both are using Brain daily and finding it
faster than their current workflow. See [Principles #8](/brain/staging/principles).

## Permission philosophy

The model is straightforward:

* **Anyone** can ask Brain to draft something.
* **Merchandisers** can approve their own drafts for most workflows.
* **Managers** can approve high-stakes things (new clients, pricing exceptions,
  manual overrides).
* **The AI agent** can never finalize without approval — but can write drafts to
  the database when they're clearly labeled as drafts (e.g., BOM drafts).

When a user tries an action they don't have permission for, Brain doesn't fail.
It prepares the draft and routes the approval to whoever does have permission.
The work continues; the gate moves to the right person.

## Related

* [Workflows](/brain/staging/workflows/index) — each workflow's permission gates
* [Principles #6](/brain/staging/principles) — permissions before action