> ## Documentation Index
> Fetch the complete documentation index at: https://help.clientx.in/llms.txt
> Use this file to discover all available pages before exploring further.
# Knowledge Base: Teach Your AI Agent About Your Product
> The ClientX knowledge base stores your content so the AI agent can answer visitor questions accurately. Add URLs, files, Q&A pairs, and objection handlers.
The knowledge base is how you teach ClientX what to say. Every piece of content you add — website pages, documents, Q\&A pairs, objection responses — gets stored as vector embeddings. When a visitor asks a question, ClientX searches those embeddings for the most relevant chunks and uses them to generate an accurate, grounded answer. The result is a conversational agent that sounds like it actually knows your product, not a generic chatbot guessing from the internet.
## What you can add
Point ClientX at your website and it automatically crawls every linked page on the same domain. Supports full-site crawls, sitemap imports, and individual URL lists. Stays fresh with automatic recrawl schedules.
Upload PDFs, Word documents, PowerPoint decks, plain text files, and Markdown. ClientX extracts the text and indexes it exactly like a crawled page.
Write specific question-and-answer pairs for product FAQs, pricing details, and support topics. Q\&A pairs take retrieval priority over crawled content when a visitor's question is a close match.
Tell the AI how to respond to common sales objections — "you're too expensive," "we already have a solution," and similar pushback. Include the ideal response and the outcome you want.
Build full battlecards for each competitor. When a visitor asks "how do you compare to X?", the agent draws on your positioning, differentiators, talk track, and proof points.
Paste any text directly — brand guidelines, policy documents, onboarding notes, or anything else you want the agent to know. Supports up to 50,000 characters per block.
## How retrieval works
When a visitor asks a question, ClientX runs a semantic search across all your knowledge sources. It finds the chunks most relevant to the question — not just keyword matches, but meaning-level similarity — and passes them as context to the AI. The agent then generates a response grounded in that specific content.
This means the quality of your answers is directly tied to the quality and coverage of your knowledge base. A well-maintained knowledge base with diverse source types produces confident, specific answers. A sparse one produces hedged, generic responses.
## The coverage score
Your knowledge base has a **coverage score** from 0 to 100 that reflects how comprehensively it covers the topics visitors ask about.
| Score | Meaning |
| ------ | -------------------------------------------- |
| 0–39 | Low — the agent will frequently lack context |
| 40–69 | Fair — most common questions are covered |
| 70–100 | Good — strong coverage across topic areas |
The score updates as you add and process new sources. You can see it at any time in the sidebar of your Knowledge Base dashboard.
## Next steps
Learn how to add URLs, files, Q\&A pairs, objections, and competitors to your knowledge base.
See which questions your agent couldn't answer and fill in the missing content.
The AI only answers from your knowledge base. It won't hallucinate facts about your product, invent pricing, or make up features. If a question falls outside what you've added, the agent will say so rather than guess — which is why regularly reviewing your gaps keeps answer quality high.