Use case
One knowledge base, askable by customers and employees
Upload your docs, crawl your website, paste your notes. Chatref builds a knowledge base anyone can query in natural language, grounded in the sources you actually wrote.
Featured on
A knowledge base use case describes turning scattered documentation into a single conversational source that both customers and employees can query in plain language. Chatref does this by ingesting three source types – uploaded files (PDFs, documents, text), crawled websites via Firecrawl, and pasted text notes – then learning from each source so it can find the relevant parts at answer time. When a question comes in, Chatref finds the most relevant parts of your content, ranks them by how well they match, and removes near-identical passages before generating a grounded response. The same knowledge base can power multiple chatbots, so you can spin up a public help-center bot and an internal team bot from one source of truth. Chatref supports 10 languages with a smart search strategy and deploys to a web widget, shareable link, iframe, or direct API. Pricing is pay-as-you-go: $50 free credit on signup, 1 to 5 credits per answer, $10 minimum topup, unlimited chatbots and unlimited team seats, with every feature unlocked for every account.
Train once · serve customers and teams · $50 free credit
At a glance
The short version.
3
Source types supported
10
Languages supported
Multiple
Bots per knowledge base
$50
Free credit on signup
- Chatref ingests three source types – uploaded files, crawled websites, and plain text – into a single source your chatbots answer from.
- Content is split into structured passages and learned so the agent can find the most relevant parts of your content for any question.
- Retrieval keeps the strongest matches, ranks them by keyword and context, and removes near-identical passages before generating the answer.
- One knowledge base can power multiple bots, so the same content serves a public help-center agent and an internal team agent without duplicating sources.
- Pay-as-you-go pricing means no seat fees, no per-bot plan tiers, and $50 of free credit to try the full experience end to end.
The problem
Static knowledge bases assume people know what to search for
Help articles, Notion pages, PDFs, and intranet posts pile up. Customers search wrong terms and give up. Employees ping Slack because finding the doc costs more than asking a colleague.
Content is there – no one can find it
The answer lives in a three-year-old help article, a product spec in Notion, and a PDF attached to an old email. Search boxes need exact keywords; your users do not have them.
Two audiences, two silos
The help center is written for customers. The internal wiki is written for staff. Both cover the same product, but the knowledge gets copied, rewritten, and drifts out of sync.
Updates never propagate in time
A policy changes, a new feature ships, a pricing tier renames. The help center catches up in a week, the internal doc in a month. Inconsistent answers make it to real customers.
How it works
From raw content to a grounded knowledge base
Add any source
Upload PDFs, documents, and text files. Paste plain text notes. Point the crawler at your website and Chatref pulls the pages through Firecrawl as clean markdown.
Build your knowledge index
Each source is split structurally into clean passages, then learned and stored so the agent can find the relevant parts of your content in milliseconds.
Find, rank, dedupe
Queries pull the strongest matching passages, then rank them by keyword match and context. Near-duplicate passages are collapsed so the agent sees the strongest signal.
Connect a bot to the base
Point a chatbot at the knowledge base, customize the system prompt and welcome message, and embed the widget on any page or share a direct link. Spin up a second bot on the same base for your internal team.
In detail
The details.
Who it serves
- Public help center and support. Embed a public bot on your marketing site or docs and let customers ask questions in their own words. Answers come from the same help articles and product pages you already maintain.
- Internal teams and new hires. Point a second bot at the same knowledge base for employee-only questions. New hires stop pinging Slack for the runbook, and support staff ramp on product details without a buddy session.
- Gated bots for partners and ops. Protect a bot with a secret token and share it with partners, ops, or a private team. Same source of truth, different audience, no duplicated content.
Compare
Chatref vs single-audience knowledge tools
Most tools pick a lane. Enterprise search tools serve employees. Help-center bots serve customers. Chatref serves both audiences from one knowledge base, priced per answer instead of per seat.
| Feature | ChatrefOur pick | Guru | Glean | Chatbase |
|---|---|---|---|---|
| Pricing model | Pay per answer | Per seat | Enterprise contract | Monthly plans |
| Seat fees | $15+ per user | Per seat | Tiered | |
| Free credit to try | $50 | Trial | Trial messages | |
| Source types | Files, website, text | 100+ connectors | 100+ connectors | Files, website, text |
| Serves customers and employees | Employees only | Employees only | Customers focus | |
| Languages supported | 10 | Multi | Multi | 80+ |
| Multiple bots per knowledge base | ||||
| Self-serve setup | ||||
| Requires enterprise sales call | For larger plans |
What counts as a knowledge base in Chatref?
A knowledge base in Chatref is a collection of sources – uploaded files, crawled websites, and pasted text – that has been learned and stored so the agent can find the relevant parts. Any chatbot you create can be pointed at that base and will answer from it.
What file formats can I upload?
Chatref accepts PDFs, documents, and plain text. The file is stored, text is extracted through Firecrawl as clean markdown, then the content flows into the same pipeline that website and text sources do.
How does answering actually work?
Every incoming question is compared against your stored content. The strongest matches are kept, ranked by keyword match and context, and near-identical passages are removed when their text overlaps more than 80 percent, so the agent answers from the relevant parts of your content.
Can one knowledge base serve both customers and my internal team?
Yes. A single knowledge base can back any number of chatbots. You can run a public customer-facing bot and an internal employee bot on the same sources, with different system prompts, welcome messages, branding, and access controls.
Does Chatref cite its sources in the answer?
Chatref answers from your own content and grounds every response in the matching passages, but the chatbot does not expose source URLs or file names in the reply today. If that is a hard requirement, reach out so we can understand your use case.
Can I update or delete sources after they are indexed?
Yes. From the knowledge base dashboard you can add, delete, or re-upload files and websites. Removing a source deletes it from the index, so the chatbot stops answering from that content immediately.
What languages does the knowledge base work in?
Answering supports 10 languages: English, Spanish, French, German, Dutch, Portuguese, Swedish, Danish, Arabic, and Japanese. A fallback retries the search in translated and keyword-only forms so queries that mix languages or contain typos still match relevant content.
Explore
Keep exploring.
Turn your docs into an askable knowledge base
Upload files, connect your website, and point any number of bots at the same source. $50 free credit on signup, no credit card required.