Your AI Agent is only as good as the information it has access to. Without a knowledge base, it relies entirely on general AI knowledge and your behavior prompt — which means it can't answer questions specific to your products, policies, or services. The Knowledge Base feature lets you feed your agent the exact information it needs to give accurate, relevant answers.
How the Knowledge Base Works
Think of the knowledge base as your AI Agent's reference library. When a customer asks a question, the agent doesn't just generate a response from thin air. Instead, it searches through your uploaded knowledge to find the most relevant information, then crafts a response based on what it finds.
This approach — called retrieval-augmented generation — gives you two major advantages:
- Accuracy — Responses are grounded in your actual content, not AI guesswork. When a customer asks about your return policy, the agent quotes your return policy, not a generic one.
- Control — You decide exactly what information the agent can reference. If a policy changes, you update the knowledge base and the agent's responses update accordingly.
Three Types of Knowledge Sources
Chatmo supports three ways to add knowledge to your agent. You can mix and match all three within the same agent.
1. Text Knowledge
The simplest and most direct method. You write or paste text directly into Chatmo, and it becomes part of your agent's knowledge.
When to use it:
- FAQs that aren't published anywhere else
- Internal guidelines you want the agent to follow
- Quick facts, pricing tables, or policy summaries
- Clarifications or corrections to information in other sources
How to add text knowledge:
- Open your AI Agent's settings and navigate to the Knowledge Base tab.
- Click Add Knowledge Source and select Text.
- Give the entry a descriptive title (e.g., "Return Policy" or "Shipping Rates - Domestic").
- Enter or paste your content in the text field.
- Click Save.
Structure your text entries clearly. Use short paragraphs, bullet points, and headers. The AI parses well-organized content more effectively than long, unbroken blocks of text. Think of each entry as a standalone reference article.
Example of well-structured text knowledge:
Code
2. Website Scraping
Point Chatmo at a URL, and it will crawl the page (or multiple pages) to extract content and add it to your agent's knowledge.
When to use it:
- You already have a comprehensive FAQ or help center on your website
- Product descriptions are published online and kept up to date
- Blog posts or guides that contain information your agent should reference
How to add website knowledge:
- In the Knowledge Base tab, click Add Knowledge Source and select Website.
- Enter the URL you want to scrape (e.g.,
https://yourcompany.com/faq). - Configure scraping options:
- Single page — Scrapes only the specified URL.
- Multi-page crawl — Follows links on the page and scrapes linked pages too (useful for help centers with multiple articles).
- Click Scrape and wait for the process to complete.
- Review the extracted content to ensure it looks correct.
- Click Save to add it to your knowledge base.
Website scraping captures the content as it exists at the time of the scrape. If you update your website content, you'll need to re-scrape the URL to keep your agent's knowledge current. Set a reminder to refresh website sources periodically — monthly is a good starting cadence.
Tips for better website scraping:
- Choose pages with clean, text-heavy content. Pages with lots of JavaScript-rendered dynamic content may not scrape cleanly.
- Avoid scraping pages behind login walls — the scraper won't be able to access them.
- After scraping, always review the extracted content. Remove any navigation menus, footers, or irrelevant content that got pulled in.
3. File Upload
Upload documents directly from your computer. This is ideal for internal resources that aren't published online.
Supported file types:
Currently, Chatmo only supports PDF files for knowledge upload.
| Format | Extension | Best For |
|---|---|---|
.pdf | Product catalogs, policy documents, manuals |
Your PDF must contain actual text content, not scanned images. If you open the PDF and cannot highlight or copy text from it, the file likely will not process correctly. PDFs generated from applications like Word, Google Docs, or other software work best. Scanned documents and image-based PDFs are not supported.
How to upload files:
- In the Knowledge Base tab, click Add Knowledge Source and select File Upload.
- Drag and drop your PDF file into the upload area, or click to browse your computer.
- Wait for the file to upload and process. Larger documents take a bit longer.
- Once processing is complete, you'll see a preview of the extracted content.
- Click Save to add it to your knowledge base.
File size limits may apply. If a document is too large, consider splitting it into smaller, topic-specific files. This often improves the AI's ability to find relevant information anyway, since smaller documents with focused content tend to produce better search results.
Best Practices for Knowledge Organization
How you organize your knowledge base matters just as much as what you put in it. A well-organized knowledge base leads to better, more accurate AI responses.
Keep entries focused. Each knowledge entry should cover one topic. Instead of one massive document covering your entire FAQ, create separate entries for "Shipping Policy," "Return Policy," "Product Sizing," etc. This helps the AI pinpoint the most relevant source for each question.
Use clear, descriptive titles. When the AI searches your knowledge base, titles help it determine relevance. "Shipping Policy - Domestic Orders" is far more useful than "Document 3."
Write for the AI, not just for humans. The AI performs best with straightforward, factual content. Avoid heavy use of idioms, sarcasm, or cultural references that could be misinterpreted. State facts clearly and directly.
Include common question phrasings. If customers frequently ask "Where's my order?" you can include that exact phrasing in your knowledge entry along with the answer. This improves matching accuracy.
Avoid contradictions. If two knowledge entries give conflicting information about the same topic, the AI may generate confused or inconsistent responses. Audit your knowledge base periodically to catch and resolve conflicts.
Updating and Removing Knowledge
Your business isn't static, and your knowledge base shouldn't be either. All knowledge sources you have added appear in a knowledge items list. You can search for items, navigate pages with pagination, and filter by content type: ALL, DOCUMENT, WEBSITE, or TEXT.
Knowledge Item Status
Each knowledge item displays a status badge indicating its processing stage:
| Status | Description |
|---|---|
| UPLOADED | The file has been uploaded and is waiting in the processing queue. |
| QUEUED | The item is queued and will be processed shortly. |
| PROCESSING | Content is being extracted and processed by the system. |
| ACTIVE | Processing is complete. The knowledge is ready for use by the AI Agent. |
| FAILED | Processing failed. Check the file or source and try again. |
Viewing a Knowledge Entry
- Navigate to the Knowledge Base tab.
- Find the entry you want to view in the knowledge items list.
- Click the View icon (eye icon) next to the entry.
- A preview modal will appear showing the knowledge content. You can use the fullscreen toggle to view the content in full screen.
Removing a Knowledge Entry
- Find the entry in the knowledge items list.
- Click the Delete icon (trash icon).
- A confirmation modal will appear asking you to confirm the deletion.
- Click Confirm to delete, or Cancel to go back.
When policies or product details change, update your knowledge base before announcing the changes to customers. Otherwise, your AI Agent might give outdated answers during the exact period when customers are most likely to ask about the changes.
Maintenance schedule suggestion:
- Weekly — Review recently added entries for accuracy.
- Monthly — Re-scrape any website sources to capture content updates.
- Quarterly — Audit the entire knowledge base. Remove outdated entries, consolidate overlapping content, and fill any gaps you've noticed from agent handoff patterns.
How the AI Uses Knowledge to Answer Questions
Understanding the mechanics helps you set up your knowledge base more effectively:
- Customer sends a question — For example: "Do you ship to Canada?"
- Semantic search — The AI doesn't just keyword-match. It understands the meaning of the question and searches your knowledge base for entries that discuss international shipping, delivery destinations, or Canada specifically.
- Relevance ranking — Multiple knowledge entries might be somewhat relevant. The AI ranks them and focuses on the most relevant matches.
- Response generation — Using the top-ranked knowledge, combined with your behavior prompt and the conversation context, the AI generates a natural response like: "Yes, we do ship to Canada! International delivery typically takes 7-14 business days, and rates are calculated at checkout based on your order weight. Just a heads up — customs and import duties are handled on the recipient's end."
- Fallback behavior — If no knowledge entry is relevant enough, the AI either gives a general response (if it can) or acknowledges that it doesn't have that information and offers to connect the customer with a human agent.
This entire process takes about two to three seconds — fast enough that the conversation feels natural.
What's Next?
- Tools Configuration — Give your agent the ability to take actions, not just answer questions.
- Testing Your AI Agent — Validate that your knowledge base is working correctly before going live.
- Create an AI Agent — Haven't set up your agent yet? Start here.