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

# Agent knowledge

> Make a team file searchable so agents and automations can pull it in as context when they work.

Any file you upload to a team can sit there as a download — or you can turn it into **Agent knowledge** so your [agents](/agents/overview) and [automations](/agents/overview) can search its contents and use it as context. This page covers turning that on, reading the sync status, and what is worth adding.

## What Agent knowledge does

When you mark a file as Agent knowledge, Erstan embeds its contents into the team's search index. From then on, agents working in that team can retrieve the relevant parts of the file when a question or task calls for it — your collections policy, a vendor master list, a chart of accounts mapping, an SOP — without you pasting it into every chat.

This is one of the two retrieval sources Erstan calls **Knowledge**: your searchable **Team Files** and Erstan-managed **NetSuite product updates**. See [Knowledge, tools, web search & models](/chat/knowledge-tools-models) for how that surfaces in chat.

<Info>
  Agent knowledge makes a file's text retrievable; it is not OCR or a document-understanding screen. For viewing a file's contents yourself, use the inline preview on the [file detail page](/content/files). Extraction from scans or images is agent behavior, not a setting you toggle here.
</Info>

## Turn a file into Agent knowledge

<Steps>
  <Step title="Open the file">
    Go to a team's **Files**, then open the file (or use its **⋯** menu). Agent knowledge is a per-file setting, so start from the file you want to make searchable.
  </Step>

  <Step title="Open Agent knowledge">
    Choose **Agent knowledge…** from the file's actions. A panel shows the **Available for knowledge** toggle and the current sync status.
  </Step>

  <Step title="Turn on Available for knowledge">
    Switch **Available for knowledge** on. Erstan queues the file for indexing.
  </Step>

  <Step title="Wait for AI search ready">
    The status moves through **Queued → Syncing → AI search ready**. Once it reads **AI search ready**, agents in that team can retrieve the file.
  </Step>
</Steps>

<Note>
  Indexing is asynchronous — a larger file takes a moment to finish. You can leave the file and come back; the status updates on its own.
</Note>

### Sync statuses

| Status              | What it means                                                   |
| ------------------- | --------------------------------------------------------------- |
| **Off**             | The file is not part of Agent knowledge and won't be retrieved. |
| **Queued**          | Indexing is scheduled and about to start.                       |
| **Syncing**         | Erstan is embedding the file's contents now.                    |
| **AI search ready** | Indexing finished; agents can search the file.                  |
| **Sync error**      | Indexing failed. Use **Retry indexing** to try again.           |

<Tip>
  A file can inherit its folder's knowledge setting instead of an explicit per-file toggle. If you keep related reference files in one folder, set knowledge at the folder level so new files added there become searchable automatically.
</Tip>

## How agents use it

Once a file is **AI search ready**, it feeds two places:

* **Chat** — when **Include Team Files** is on in the composer's Knowledge menu, Erstan can pull relevant passages from your searchable files into a reply. Scope is set there (this team or all your teams). See [Knowledge, tools, web search & models](/chat/knowledge-tools-models).
* **Agents and automations** — an agent that has Knowledge available can retrieve from these files during a [run](/agents/run-and-tasks). NetSuite release and feature questions can separately use Erstan-managed product updates.

You don't have to name the file in your message. Erstan matches your question against the indexed content and brings back the most relevant parts.

## What to add

Add reference material that agents should apply consistently, and keep it tidy so retrieval stays sharp.

<AccordionGroup>
  <Accordion title="Good candidates">
    * Policies and SOPs — collections cadence, write-off thresholds, approval rules.
    * Reference data — vendor lists, customer segments, GL/account mappings, code lists.
    * Templates and standards — dunning-email wording, report formats, naming conventions.
    * Background context — close checklists, period notes, team glossaries.
  </Accordion>

  <Accordion title="Keep in mind">
    * **Text wins.** Text, Markdown, and spreadsheet content index well. Scanned images and binary files carry less retrievable text.
    * **Currency matters.** Update or replace files when the underlying policy changes — agents retrieve what's indexed, not a newer copy you haven't added.
    * **Stay scoped.** Add files to the team whose agents need them. Knowledge is team-scoped, and chat search can be limited to one team.
    * **Size limits.** Uploads are capped at 10 MB per file and 25 files per upload (a NetSuite File Cabinet-safe limit). Split very large references if needed.
  </Accordion>
</AccordionGroup>

<Warning>
  Marking a file as Agent knowledge makes its contents retrievable by anyone whose agents run in that team. Don't add files with secrets or data that shouldn't be surfaced in replies. Turning **Available for knowledge** off removes the file from future retrieval.
</Warning>

## Next steps

<CardGroup cols={2}>
  <Card title="Files: upload, preview & edit" icon="file" href="/content/files">
    Get files into a team, preview them, and edit text-based files in the browser.
  </Card>

  <Card title="Knowledge, tools & models" icon="book-open" href="/chat/knowledge-tools-models">
    Search Team Files in chat and understand automatic NetSuite product-update retrieval.
  </Card>

  <Card title="Giving agents tools & skills" icon="wrench" href="/agents/tools-skills-connections">
    Pair knowledge with the skills and connectors an agent needs to act.
  </Card>

  <Card title="Documents and the editor" icon="pen-line" href="/content/documents">
    Author rich documents that agents can read and write alongside you.
  </Card>
</CardGroup>