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

# Knowledge Sources

Knowledge sources are the content your agent draws answers from. The more relevant and well-structured your sources are, the better your agent performs. You'll find them under **Knowledge** in the sidebar.

> **Access:** Adding, editing, and deleting sources requires Editor or Owner access. Viewers can browse the list but can't make changes.

***

## Source types

Klaro supports three ways to feed content to your agent:

| Type         | Best for                                                                                  |
| ------------ | ----------------------------------------------------------------------------------------- |
| **Websites** | Product pages, help centers, FAQs, blog posts - anything publicly accessible by URL       |
| **PDFs**     | Documentation, manuals, whitepapers, anything that lives in a file                        |
| **Text**     | Short reference content, internal FAQs, policies - anything you want to paste in directly |

Each source type is indexed separately but queried together when a visitor asks a question.

***

## Status lifecycle

After you add a source, Klaro processes it in the background. You can watch the status update in real time - no need to refresh.

### Website

```text theme={null}
pending → crawling → indexing → completed → failed
```

* **Pending:** source created, waiting to be picked up
* **Crawling:** Klaro is fetching the page content from the URL
* **Indexing:** content has been retrieved and is being split into chunks and embedded
* **Completed:** source is live and searchable
* **Failed:** something went wrong at one of the steps (bad URL, unreachable page, etc.)

### PDF

```text theme={null}
pending → indexing → completed → failed
```

PDFs skip the crawling step - once uploaded, they go straight to indexing.

### Text

```text theme={null}
indexing → completed → failed
```

Text sources start indexing immediately on creation.

> **If a source shows "failed":** the most common reasons are a URL that requires login, a PDF that's too large or too long, or a text source that exceeded the word limit. Fix the issue and re-add the source (or re-scrape for websites).

***

## Adding knowledge sources

### Website

You have three ways to add website content:

**Single** - paste one URL and submit. Good for a specific page you know you want indexed.

**Multiple** - paste up to **10 URLs**, one per line. Use this when you have a list of pages ready to go.

**Crawl** - enter a base URL and Klaro maps the site, showing you up to 30 discovered pages. Pick the ones you want (up to 10 at a time) and add them in one go. This is the most efficient way to index an existing help center or documentation site.

URLs are normalized before saving - you can paste with or without `https://` and trailing slashes, Klaro handles it.

### PDF

Click **+ Add Document**, then drag and drop your file or click to browse. One file per upload.

Constraints:

* Format: PDF only
* Max size: **10 MB**
* Max pages: **50 pages**

The title defaults to the filename (without `.pdf`). You can't rename it after upload - if you need a different title, delete and re-upload.

### Text

Click **Add Text Source**, enter a title and paste your content. The title must be unique per agent - adding two text sources with the same title will fail.

Constraints:

* Max: **10,000 words**
* A live word counter shows your current count as you type; the submit button disables when you go over

Text sources are useful for internal knowledge that isn't published anywhere - things like escalation procedures, pricing tiers, or product positioning notes that you want the agent to know but don't live on your website.

***

## Re-scraping websites

Websites can be manually re-scraped to pick up updated content. Find the source in the list and click **Re-scrape**.

A few things to know:

* Re-scraping counts against your **monthly website scraping limit** (same as adding a new website)
* You can't re-scrape while a source is currently being crawled or indexed - the button is disabled during processing
* When a re-scrape completes, the old content is fully replaced - not merged

PDFs and text sources don't have a re-index option. If the content changes, delete the source and re-add it.

***

## Deleting sources

You can delete any source from the list. Deletion is permanent and immediate - the source is removed from the index and your agent will no longer use it for answers

***

## Plan limits

Two limits apply to knowledge sources, and both are scoped to your workspace:

**Total knowledge sources** - the combined count of all source types across all your agents. Once you hit the limit, you'll need to delete existing sources or upgrade your plan to add more.

**Website scrapes per month** - counts both new websites added and re-scrapes of existing ones. Resets at the start of each billing month.

| Plan    | Max knowledge sources | Max website scrapes/month |
| ------- | --------------------- | ------------------------- |
| Free    | 5                     | 5                         |
| Starter | 25                    | 25                        |
| Builder | 100                   | 150                       |
| Pro     | 400                   | 500                       |

You can see your current usage directly on the Knowledge Sources page. If you hit a limit mid-workflow, you'll see an inline error message with your current counts - head to **Settings → Billing** to review your plan.

***

**Next up:** [Configure → Customize](../en/configure/customize) - enable human escalation and lead capture.