Feature 02 · Answers from your docs
Turn your docs into accurate, on-brand answers.
Paste your knowledge base or a system prompt and the bot answers from it — staying on topic, and saying so when it doesn't know.
Generic chatbots guess, ramble, and confidently invent answers — then your team cleans up the mess.
Your documentation already holds the answers to most support questions. Clanker Support grounds every reply in the knowledge you give it, so customers get accurate, on-brand answers instead of generic AI filler.
Just as important: it admits the limits of what it knows. Rather than inventing a confident wrong answer, it defers — and hands off to a human when it should.
What you get
Grounded in your content
Paste a knowledge base or write a system prompt. The bot answers from that source of truth, not the open internet.
Stays on topic
It sticks to what you've told it about your product and politely declines to wander into unrelated territory.
Knows when to stop
When the docs don't cover a question, it says so and escalates — no confident hallucinations.
What it means in practice
- ✓Update your knowledge base in the dashboard and answers change immediately — no retraining.
- ✓Tune tone and boundaries with a plain-language system prompt, per project.
- ✓Pair it with the escalation threshold so uncertain answers become human hand-offs.
Explore the rest
Drop-in widget
One script tag. Loads in a shadow DOM so your styles never leak — and theirs never leak into you.
03Escalates to humans
When the bot can't help, the conversation lands in your inbox with full context — no lost threads.
04Email threading
Replies go out as email and customer responses thread back into the same conversation automatically.
05Any model, any time
Built on LLM Gateway. Swap GPT, Claude, or a custom model per project — a config change, not a rewrite.
06Self-hostable
Open architecture on serverless edge infra. Run the whole stack on your own account. No surprise vendors.
Frequently asked questions
How does the bot know about my product?
You give it the knowledge. Paste your knowledge base or write a system prompt in project settings, and the bot answers from that source of truth — not the open internet. Update the content in the dashboard and answers change immediately, with no retraining.
What happens when the bot doesn't know an answer?
It admits the limit instead of inventing one. When your docs don't cover a question, the bot says so and escalates to a human rather than producing a confident wrong answer — so customers aren't misled and your team isn't cleaning up after it.
Can I control the bot's tone and boundaries?
Yes. A plain-language system prompt sets the bot's tone and what it should and shouldn't talk about, per project. It keeps replies on-brand and on-topic, and politely declines to wander into unrelated territory.
Ship support today