> For the complete documentation index, see [llms.txt](https://docs.zipchat.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zipchat.ai/ai-training-and-rules/knowledge-base.md).

# Knowledge Base Management

Build your Zipchat AI agent's knowledge base from web pages, sitemaps, product feeds, PDFs, texts, codebases and databases, and keep it fresh.

The knowledge base is what your AI agent knows: your products, policies and help content. Zipchat builds it from your website and from the sources you add.

![Knowledge base in Zipchat](https://2043605213-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FvYpZDb6HsnLHFu8ZygCS%2Fuploads%2Fgit-blob-d8488fff2e7f3c8289af6091095f4cfcb70ae576%2Fknowledge-base-2026.png?alt=media)

## Add sources

Go to **Knowledge → Knowledge Base** and click **Add new source**. You can add:

| Source                                | Use it for                                                                              |
| ------------------------------------- | --------------------------------------------------------------------------------------- |
| **Single page**                       | One URL, such as a policy or FAQ page.                                                  |
| **Multi-page website**                | Your whole site or a section of it, crawled from a starting URL.                        |
| **PDF**                               | Manuals, spec sheets, size guides or policies.                                          |
| **Sitemap (XML)**                     | Many pages at once, from your sitemap.                                                  |
| **List of Links (CSV)**               | A chosen list of URLs.                                                                  |
| **Custom text**                       | Information that is not on your website. The agent can share the links you write in it. |
| **Google Merchant Feed (XML or CSV)** | Complete product data: titles, prices, images and availability.                         |
| **Codebase** and **Database**         | Technical sources, covered at the [end of this page](#codebase-and-database-sources).   |

Zipchat also imports your products, policies, pages and blog posts from your website.

To bring in an existing FAQ or help center, add it as a **Multi-page website** or **Sitemap (XML)**, or list its article URLs in a **List of Links (CSV)**.

The list of sources has filters by type, a search box and a status filter, so you can find a page and check that it was saved.

## Live product data on Shopify

On Shopify stores, the agent uses live product data for price and availability. When crawled page text disagrees with live data, live data wins.

## Keeping it fresh

Pages are refreshed automatically on a schedule: daily, weekly or weekly dispersed. You can also refresh a single page from the list.

## Page limit

Each plan includes a number of knowledge base pages. Every item counts as one page: a web page, a product, a feed item, a PDF or a text. Imported products and pages count too.

| Plan       | Knowledge base pages |
| ---------- | -------------------- |
| Starter    | 1,000                |
| Pro        | 75,000               |
| Scale      | 300,000              |
| Scale Plus | Unlimited            |
| Unlimited  | Unlimited            |

When you reach the limit, new pages are not added until you upgrade or remove pages you don't need, such as tag or archive pages. See [Subscription](https://docs.zipchat.ai/account-and-billing/subscription.md).

## Out-of-stock products and page clutter

In **Channels → Configuration → Crawling settings** you can exclude out-of-stock products and strip headers, footers and other repeated elements. See [Crawling Settings](https://docs.zipchat.ai/settings-tracking-and-security/crawling-settings.md).

## Corrections and localisations

The same section has two more tabs:

* **Corrections:** teach the agent how to answer specific questions. The number next to the tab shows how many customer questions are waiting for an answer. See [Corrections](https://docs.zipchat.ai/ai-training-and-rules/corrections.md).
* **Localisations:** use different pages and texts per market, by URL pattern.

## Codebase and Database sources

The last two options in **Add new source** connect the agent to technical data. Every account has them. In the list of sources, the **Codebases** and **Databases** filters show them.

### Codebase

Enter the **Git URL** of a repository. Zipchat reads it, so the agent can answer questions based on your code. This suits software products, apps and custom storefronts.

Zipchat supports GitHub, GitLab and Bitbucket repositories, including private ones. You connect with the repository URL and an access token; Bitbucket also needs a username. Zipchat checks access before it connects and stores the token encrypted. Use a token that can only read the repository.

### Database

1. Enter the **Host**, **Port**, **Database name**, **Username** and **Password**.
2. Click **Verify database access**.
3. Click **Add new source**.

Zipchat supports PostgreSQL databases. It only runs read queries (SELECT) on your database. Each query stops after 10 seconds and returns at most 100 rows. Zipchat stores the password encrypted. Create a dedicated read-only database user for this connection.

## FAQ

### The agent uses an old price or description. What do I do?

Refresh the page in the list. On Shopify, live product data is used for price and availability.

### How many knowledge base pages does my plan include?

Starter includes 1,000 pages, Pro 75,000, Scale 300,000, and Scale Plus and Unlimited have no limit. Products count as pages.

### Why are my pages not added?

You may have reached your page limit, or your website blocks the crawler. On Cloudflare, follow [Allow Zipchat to crawl your website](https://docs.zipchat.ai/settings-tracking-and-security/crawling-settings/cloudflare-allow-zipchat-to-crawl-your-webiste.md). For other firewalls, contact <support@zipchat.ai>.

### Which databases can I connect?

PostgreSQL. Zipchat only runs read queries, with a 10-second limit and up to 100 rows per query.

### Can Zipchat change data in my database?

No. Zipchat only runs read queries. Use a read-only database user for extra safety.

### Do knowledge sources use my AI replies?

No. Only replies written by the agent count toward your plan.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.zipchat.ai/ai-training-and-rules/knowledge-base.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
