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

# How VirtualBrain Works

> Three building blocks, one engine: the mental model behind every feature

VirtualBrain is built on three core pillars: **Brains** to centralize knowledge, **Chats** to question it in conversation, and **Apps** to turn entire workflows into repeatable automations. Together they let a firm encode how it works once, then run that expertise on every engagement.

## The three pillars

<CardGroup cols={3}>
  <Card title="Brains" icon="brain" href="/features/brains/overview">
    A secure, private space for a related set of documents, built around a client, a project, a team, or your whole firm. VirtualBrain reads and indexes everything you add, and you choose who it's shared with.
  </Card>

  <Card title="Chats" icon="message" href="/features/chats/overview">
    Ask questions in plain language and get answers built from your documents (and the web when you choose), with every claim traced to its source.
  </Card>

  <Card title="Apps" icon="grid-2" href="/features/apps/overview">
    A repeatable, multi-step workflow that turns inputs into a finished deliverable (a due-diligence summary, a proposal, a memo) in a consistent format every time.
  </Card>
</CardGroup>

<Info>
  **The relationship in one line:** you put documents into a Brain, you explore and question them in Chat, and you industrialize the work you repeat as an App.
</Info>

## How a request flows

Behind the chat sits an orchestration engine, the **Mainframe**. Rather than following a fixed script, it reasons about what your question needs, then uses its tools to gather the context required to answer it, looping until it has enough.

<Steps>
  <Step title="You ask a question or launch an App">
    The request is grounded in the Brain(s) you select, plus the web if you switch it on.
  </Step>

  <Step title="The Mainframe gathers the context it needs">
    It calls on whichever of its four capabilities the question requires: retrieval finds the passages that bear on it, web search brings in online information, spreadsheet analysis runs real code to get the numbers right, and document comparison reports what changed between two versions. See [the Mainframe](/features/mainframe).
  </Step>

  <Step title="A final agent builds the deliverable">
    It combines everything the Mainframe gathered into one well-structured answer, with a citation on every claim, ready for you to verify.
  </Step>
</Steps>

The Mainframe works the way a senior partner does, delegating to specialists and checking their work. You can watch each tool report back as the answer is researched; see [Sourced Answers](/features/chats/sourced-answers).

## Where to go next

<CardGroup cols={2}>
  <Card title="Create your first Brain" icon="folder-plus" href="/features/brains/creating-brains">
    Upload documents, sync cloud storage, or capture the web.
  </Card>

  <Card title="Ask your first question" icon="comments" href="/features/chats/asking-questions">
    Ground a chat in your Brains and get a sourced answer.
  </Card>

  <Card title="Run a pre-built App" icon="play" href="/features/apps/using-templates">
    Start from 50+ templates built around professional use cases.
  </Card>

  <Card title="Set up your workspace" icon="gear" href="/features/admin-dashboard/overview">
    Teams, shared prompts, integration policy, usage, and templates.
  </Card>
</CardGroup>