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

# Add knowledge to your agent

> Teach a Jaaxis agent with text, Q&A, website pages, and PDF, DOCX, XLSX, CSV, or TXT files.

Knowledge gives your agent the business facts it may use in customer replies.

Open **Agents**, choose an agent, select **Training**, and use the **Knowledge** tabs.

## Choose a source type

<Tabs>
  <Tab title="Text">
    Paste a focused policy, product description, service explanation, or instruction. Give each entry one clear topic so it is easy to maintain.
  </Tab>

  <Tab title="Q&A">
    Add a real customer question and its approved answer. Q\&A is useful for wording that must be handled directly.
  </Tab>

  <Tab title="Website">
    Add one complete public URL, or choose the option to discover pages from a website root. Page discovery lets you select up to 50 pages before importing.
  </Tab>

  <Tab title="Files">
    Upload PDF, DOCX, XLSX, CSV, or TXT files. Each file must be 15 MB or smaller. You can select more than one file.
  </Tab>
</Tabs>

## Write knowledge that retrieves well

* Put one policy or subject in each entry.
* Use the same names customers use for products and services.
* Include conditions, exceptions, dates, and regions when they change the answer.
* Remove navigation, legal boilerplate, and unrelated text from pasted content.
* Never upload credentials, private keys, payment details, or data that customers should not receive.

## Optional availability dates

Use the schedule controls when content should only apply during a campaign or season. The entry can have a start time, an end time, or both.

## After adding content

New sources appear as processing before they are ready. Open **Trained** and wait until processing finishes, then [test representative questions in Preview](/en/build/preview).

<Note>
  A successful upload only means Jaaxis accepted the source. Verify the resulting answers before using the agent with customers.
</Note>