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

# Add knowledge to your agent

> Train an agent with websites, files, text, and question-and-answer pairs.

Knowledge sources give your agent the product and policy information it needs to answer customers. Add focused, current content and test it before publishing changes.

## Choose a source type

Each source type fits a different content workflow:

* **Website**: index a site, a single page, or a batch of URLs
* **Files**: upload existing support or product documents
* **Text**: paste content that doesn't have a public URL
* **Q\&A**: write an approved answer for a specific question

<Steps>
  <Step title="Open Sources">
    Select your agent, then select **Sources** in the sidebar. Choose the source type you want to
    add.
  </Step>

  <Step title="Add the content">
    Enter a title and the requested content. For files, choose the files from your device. For a
    website batch, add one URL per line.
  </Step>

  <Step title="Process the source">
    Create the source and start processing when prompted. Website indexing can continue in the
    background.
  </Step>

  <Step title="Check the result">
    Open the source to review its status and preview. If processing fails, correct the content or
    URL and process it again.
  </Step>
</Steps>

## Keep website content current

Open a website source to configure retraining. Retraining availability depends on your plan. You can also start processing manually when the source needs an immediate update.

<Warning>
  Archiving removes a source from the agent's available knowledge. Test the agent after archiving
  content that may support existing answers.
</Warning>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.