A chat agent for your website that answers from what you tell it.
You write what your business does. It answers visitors' questions about that, in your voice, and stays on the subject. One line of HTML on your page — no plugin, no framework, nothing to host.
Priced by the plan, not by the message. No per-message rate, no credits, no overage.
No card for the trial, and no invite needed — sign up and you are answering visitors in about ten minutes.
What it is
A first-contact agent, not a search box
Most visitors arrive with the same handful of questions: what do you do, do you cover my case, how much, how do I reach a person. Answering those is not hard — it is just relentless, and it happens at the hours you are not there.
kavilo answers them from Context you write — or from public pages you copy into Context at setup. It is not trained on your data, and while it is answering it cannot fetch a page. Tell it to say when the answer is missing and offer a way to contact you, rather than inventing an answer that you then have to honour.
How it works
Three things, in order
1. Describe the business
Four fields: Greeting (the opening line), Persona (who it is and how it speaks), Context (what is true — services, hours, prices, policies), and Guardrails (what it must never say or promise). Plain prose. No flow charts, intent trees or training run.
2. Paste one line into your page
That is the whole integration. It works on Webflow, Squarespace, WordPress, Shopify, a hand-written HTML file — anywhere you can add a script tag. You also tell us which domains may use your key, so another website cannot use it through a visitor's browser.
3. Read what people asked
Conversations can be read in the portal, and a transcript can be emailed to you when a chat ends. The questions visitors actually ask tend to be worth more than the answers.
Pricing
Flat, with no overage
We price by the plan, not by the message. No per-message rate, no credits, no overage.
| Plan | Per month | Chat bots | Messages / month | Context / bot | Badge | Live data (MCP) |
|---|---|---|---|---|---|---|
| lite | $9 | 1 | 1,000 | 20,000 characters | Shown | — |
| starter | $29 | 1 | 4,000 | 40,000 characters | Not shown | — |
| pro | $79 | 5 | 15,000 | 80,000 characters | Not shown | — |
| business | $199 | 25 | 50,000 | 100,000 characters | Not shown | — |
| enterprise | $499 | 25 | 150,000 | 100,000 characters | Not shown | Yes |
New accounts start with 30 days at Starter's message, Context and bot allowances, no card. The trial keeps the kavilo badge and does not include custom CSS. When it ends, replies pause until you choose a paid plan. If you cancel, your plan remains active through the period already paid for and replies then pause.
Reach the daily safety ceiling or monthly allowance and replies pause instead of generating an overage invoice. Monthly counts reset at the start of the UTC calendar month.
If hosting it yourself matters
The same agent, in your network
Some questions cannot be answered from a description because the answer is in a live system. For those, the Enterprise plan lets the agent call one tool on an MCP server you run — your own API, database bridge, or another service that speaks the protocol — and answer from what it returns. Your source system stays in your network; the request and the tool's response pass through our hosted service at question time.
If your data cannot leave your network at all, the same agent also runs fully inside your own infrastructure, next to the data, reachable only from your intranet. With local inference nothing leaves — not the schema, not the queries, not the rows. That is a different posture from this hosted service, and we would rather say so plainly than sell the hosted one as equivalent.