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

# Knowledge Bases

> Give agents long-term, searchable memory over your unstructured documents — brand guidelines, internal wikis, customer research, product docs, anything.

A **Knowledge Base** is a searchable index of unstructured content — PDFs, markdown docs, text files, scraped pages, transcripts. Agents query knowledge bases when they need authoritative context that's specific to your business.

## When to use this

Use a knowledge base when you have a body of content that doesn't change minute-to-minute but matters for a lot of agent work:

* **Brand voice and style guides** — every piece of content the agent generates should follow these
* **Internal wikis and SOPs** — how your team does things
* **Customer research and persona docs** — so agents can ground recommendations in your actual ICP
* **Product documentation** — so support and sales agents can answer questions accurately
* **Past campaign post-mortems** — so the next campaign benefits from what you learned

For *fast-changing* structured data (records, rows, time series), use [Hyper Database](/data/hyper-database) instead. For *individual files* you want an agent to use once, just attach the file directly in chat or via [Files](/data/files).

## Creating a knowledge base

<Steps>
  <Step title="Open Knowledge Bases">
    From the sidebar, click **More** → **Knowledge Bases** → **Create knowledge base**.
  </Step>

  <Step title="Name and describe it">
    Give it a clear name (e.g. "Brand Guidelines", "Product Wiki", "ICP Research") and a description telling agents *when* to use it. The description matters — agents read it before deciding to query.
  </Step>

  <Step title="Add content">
    Upload files (PDFs, markdown, txt, docx, csv) and/or paste URLs to scrape. Hyper indexes everything for semantic search.
  </Step>

  <Step title="Attach to an agent">
    Open an agent's **Resources** tab and add the knowledge base. From now on the agent will query it when it's relevant to the task.
  </Step>
</Steps>

## How agents use knowledge bases

When an agent has a knowledge base attached, it can:

* **Search** — "Find what our brand guidelines say about tone for paid ads"
* **Cite** — return specific passages so you can verify the source
* **Compose** — pull excerpts into reports, drafts, and analyses

Agents only query the knowledge base when it's relevant — they don't dump the whole index into every chat. The description you wrote when creating it tells the agent when to reach for it.

## Updating content

Knowledge bases are meant to live and grow. Add new files anytime; remove or update old ones. Re-indexing is automatic — content is searchable within minutes of being added.

If the underlying content changes a lot (e.g. live product docs), consider:

* Re-uploading files on a schedule via a [task](/agents#tasks)
* Pointing a scraper task at the source URLs and refreshing weekly

## Going further

* For structured data, use [Hyper Database](/data/hyper-database)
* To turn knowledge-base outputs into shared dashboards, see [Interfaces](/data/interfaces)
* To attach a single file to a single chat instead of creating a knowledge base, see [Files](/data/files)