It answers what your FAQ already answers. It stops where your FAQ stops.
Feed it your pages, your PDFs, your price list and your FAQs. It replies in your voice, shows which source each answer came from, and hands the conversation to a person when it doesn’t know — instead of inventing something you have to apologise for.
Free plan · no card needed · cancel any time
02:14
are you open on sunday and do you do balayage
We’re open Sunday 10–6, and yes — balayage starts at ₹3,500. Source: Services page.
can you do it cheaper
That one’s for a human — I’ve passed it to the team and they’ll reply in the morning.
An illustration of a ReplyKing flow, not a real customer conversation.
The twelve questions you answer every single day
- “What are your timings?” “Do you deliver here?” “How much for the small one?” — again, at 2am, from three people.
- You tried a chatbot once. It confidently told a customer something that wasn’t true, and you turned it off.
- The person who knows the answers is you, and you are in a meeting until six.
What happens instead
Four steps, and every one of them is a real setting
- 1
Give it something to read
A website URL, pasted text, a PDF or a list of question-and-answer pairs. It is re-indexed when you change it, and each version is kept, so you always know what it was reading when it said something.
- 2
Tell it who it is and what it may not do
A persona in your words, and guardrails it holds to: what it never discusses, when it must hand over, how it signs off.
- 3
Try it before anyone else does
A playground where you ask it the awkward questions yourself and see the sources it used, before it speaks to a single customer.
- 4
Let it out, on a leash
Turn it on for one step in a flow, or let autopilot answer only the chats nobody picked up — with a daily cap, a limit on replies in a row, and a pause the moment a human joins.
What’s inside
What’s inside
The AI features we shipped, not the ones on a roadmap slide.
Grounded answers, with citations
It answers from your knowledge base and says which document the answer came from. When the base doesn’t cover the question, it says so and hands over.
Off by default
Autopilot is something you switch on, per workspace, with a mode you choose, a daily cap, a limit on how many times in a row it may reply, and a note to the team when it hands over.
A step, when you want one
An AI reply or an intent check can be one step inside a flow, so most of the automation stays deterministic and only the part that needs judgement is left to the model.
Tools it has to be given
Anything the agent can do beyond talking is a tool you grant, some of them needing approval before they run, and every use is written to an audit trail.
Credits, metered and visible
AI usage is counted in credits included with your plan. At 80% we email; at 100% it pauses until the 1st or a top-up. There is no overage charge.
A draft for a human
In the inbox it can suggest the reply instead of sending it — useful while you are still deciding how much you trust it.
Keep reading
The rest of what this does
FAQ
Questions people ask first
Let it answer the 2am questions
Every plan includes AI credits. Give it your FAQ page, ask it a few awkward questions in the playground, and decide from there.
Start free
