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

# Brain — org knowledge

> Documents, web captures, and memories available to your organization's agents.

Ask a new teammate to write a customer email on their first day and
you'll get something generic — they don't yet know how your company
sounds, who you sell to, or what you've already promised. Your agents
start the same way. The **Brain** is the shared memory that fixes
that: write down how the company works once, and every agent draws on
it from then on — instead of you re-explaining the same context on
every task.

## Why it exists

AI agents are only as good as the **context** they receive. If
Athena knows nothing about how your company talks to customers, she
sounds generic. If Marcus doesn't have access to the pitch deck your
team created, he makes up arguments from thin air.

The **Brain** is where that context lives — curated documents, captures
of relevant pages, the org's voice\_md, and memories accumulated from
past conversations. The same store feeds Athena, the operator who runs
your org, the specialists working under her, and — scoped privately to
you — your Digital Twin. Everything is organization-scoped and versioned.

## What you store in the Brain

<CardGroup cols={2}>
  <Card title="Documents" icon="file-lines">
    PDFs, markdowns, links, transcripts. Typically: pitch deck,
    defined ICP, brand voice, internal FAQ, product manual.
  </Card>

  <Card title="Knowledge captures" icon="bookmark">
    Snippets clipped from web pages during operation. Each
    capture saves the source URL + excerpt + a note from the person who captured it.
  </Card>

  <Card title="Voice_md" icon="comment">
    Markdown describing the **org's voice** — tone, vocabulary,
    signature, channel-specific greeting policy. Edited by the team,
    versioned, audited.
  </Card>

  <Card title="Agent memories" icon="brain">
    Knowledge each agent accumulates over time — learned facts,
    preferences, past decisions. Not a document you edit; it's the
    result of operation.
  </Card>
</CardGroup>

## How agents query the Brain

No agent works from a blank page. Every agent that needs context runs
a **semantic search** against the Brain — not exact text matching, but
"is this snippet relevant to what I'm doing right now?"

Examples:

* **Marcus** composing an email → pulls the `voice_md` + the pitch deck
  * recent captures about the lead in question
* **Athena** answering an internal question → searches docs that cover
  that topic, cites the sources
* **Scout** researching a lead → cross-references the public findings with
  the Brain's ICP to assess fit

## Document upload

<Steps>
  <Step title="Settings → Brain → Upload">
    Drag PDFs / MDs / text files. Or paste URLs that Apollo Space
    will fetch and index.
  </Step>

  <Step title="Apollo Space extracts + chunks">
    Text is extracted and split into pieces sized for semantic
    search.
  </Step>

  <Step title="Embeddings run in the background">
    Each chunk becomes a numerical vector representing its meaning.
    When an agent searches for "brand voice," the system finds
    semantically similar chunks — not just ones with the exact phrase.
  </Step>

  <Step title="Available to agents">
    Once embedding completes (seconds to a few minutes), every
    agent in the org can query it.
  </Step>
</Steps>

## Knowledge captures (clipping)

While browsing in Apollo Space (e.g., while Scout is researching,
or while you're reading an article opened via `read_url`), you can
**capture** excerpts with a single click.

Each capture stores:

* Source URL
* Selected excerpt (text)
* Optional free-form note from you
* Who captured it + when

Captures become part of the Brain — agents find them in semantic
searches from that point on.

## Permissions — org-wide vs personal

Apollo Space's Brain has two scopes:

<CardGroup cols={2}>
  <Card title="Org-wide" icon="globe">
    Visible to all members + all agents. Use for:
    voice\_md, institutional deck, ICP, corporate FAQ, captures
    relevant to the whole team.
  </Card>

  <Card title="Personal (per-user)" icon="user-lock">
    Visible only to you + agents operating "on your behalf"
    (including your [Digital Twin](/en/agents/digital-twins)). Use for:
    personal notes, context that only makes sense to you.
  </Card>
</CardGroup>

The separation is **fenced at the database level** — another org member
cannot see your personal brain even if they search manually.

## Search: how it works in practice

You can search the Brain manually too — it's not just for agents:

* **Direct search**: search field → type what you want to find
* **Filters**: by type (doc / capture / voice), by date, by author
* **Results ranked** by semantic relevance, with the source highlighted

Useful for "where did I see that information about client X?" three
weeks later.

## Limits and retention

* **Document size**: individual files up to 100 MB. For larger
  documents, split them or contact support.
* **Org total**: a generous limit that scales with your plan —
  details under **Settings → Billing**.
* **Retention**: documents remain **as long as the org exists**. If you
  cancel the account, the Brain is deleted after the grace period (90
  days by default).
* **Export**: you can export the entire Brain at any time
  (Settings → Brain → Export). Format: ZIP with original files +
  JSON manifest.

## Next steps

<CardGroup cols={2}>
  <Card title="Athena — the main user" icon="user-tie" href="/en/agents/athena">
    Athena queries the Brain on every operation.
  </Card>

  <Card title="Marcus — uses the voice_md" icon="paper-plane" href="/en/agents/marcus">
    Marcus loads the voice\_md for every composition.
  </Card>

  <Card title="Trust — isolation" icon="shield-check" href="/en/trust/multi-tenant">
    How the Brain is isolated between orgs.
  </Card>
</CardGroup>