How to read the result
The calculator shows two costs that you pay to two different companies. The AI cost is what your model provider charges for tokens. The platform cost is the SiteTrellis plan. Both appear in the total so you can compare a bring-your-own-key setup with a bundled chatbot quote on equal terms.
- Conversations per month. One conversation is one visitor session with at least one visitor message. Page views and widgets that were opened but never used do not count.
- Visitor messages per conversation. The bigger this is, the more each new message costs, because the whole conversation so far is sent to the model again every turn.
- Prompt and knowledge tokens. The instructions and facts you paste into the agent are sent with every message. A long knowledge block multiplies across all of them. Open “Advanced assumptions” to change it.
- Reply length. Output tokens are priced higher than input tokens at every provider, but chat replies are short.
If you do not know your traffic yet, start with a modest number, then check the dashboard after a few weeks and adjust. The Free plan’s allowance is a safe place to measure.
Methodology and assumptions
The maths is deliberately simple so you can check it by hand. For a conversation of n visitor messages, every turn sends the system prompt and knowledge (the base tokens), the new visitor message and the whole history so far, and receives one reply.
input tokens = n x (base + visitor) + (visitor + reply) x n x (n - 1) / 2
output tokens = n x reply
cost = (input x input price + output x output price) / 1,000,000The defaults are 6 visitor messages per conversation, 1,500 tokens of prompt and knowledge sent every turn, and 180-token replies, with a visitor message of about 40 tokens. Model prices are the list prices in our model catalog, in US dollars per million tokens. Three simplifications make the estimate conservative for some setups and optimistic for others:
- No prompt caching. Several providers discount repeated prompt text. If you use it, real input costs can be lower.
- No tool calls. A custom HTTP tool call adds tokens for the call and its result.
- No long-context surcharges beyond the catalog price, and no taxes or currency conversion.
Prices marked with an asterisk have not yet been confirmed against the provider’s own pricing page. OpenRouter lists other labs’ models at pass-through prices, so it is not shown separately. Prices change, so treat any figure here as a starting point and confirm it with your provider.
This tool never sends your numbers anywhere. The maths runs in your browser, and the first result is rendered by the server so the page works without JavaScript.
The SiteTrellis plans behind the “smallest plan that fits” line
The plan suggestion picks the smallest plan whose monthly conversation allowance covers your number. Plans are read from the same table as the pricing page, so they always agree with it.
| Plan | Price per month | Conversations per month | Websites | Agents | Fallback steps per agent |
|---|---|---|---|---|---|
| Free | $0 | 100 | 1 | 1 | 2 |
| Starter | $19 | 1,000 | 1 | 2 | 3 |
| Pro | $59 | 5,000 | 5 | 10 | 5 |
| Agency | $149 | 20,000 | 25 | 50 | 5 |
Plans cover the platform only. Your AI provider bills the tokens. Full plan comparison.
The Free plan is a hard monthly cap, while paid plans have a soft cap with a buffer, and chats already in progress are never interrupted. See pricing for every limit.
Worked examples from the catalog
To make the numbers concrete, here is the AI cost of 1,000 conversations under the default assumptions. These are estimates computed from the catalog prices, not promises.
- Claude Haiku 5.5 (Anthropic): about $1.79 per 1,000 conversations.
- GPT-6 Luna (OpenAI): about $1.79 per 1,000 conversations.
- Claude Sonnet 5.5 (Anthropic): about $35.88 per 1,000 conversations.
- Claude Fable 5.1 (Anthropic): about $179 per 1,000 conversations.
The spread between a small and a large model is the main lever you control, far bigger than anything else on this page. For most support and lead-capture chats, a smaller model with good instructions and accurate knowledge is a sensible place to start; test two models side by side in the playground before you decide. Our guide on choosing an LLM for website chat goes deeper.
Using the calculator to check other chatbot quotes
Chatbot pricing comes in different units, which makes quotes hard to compare. Credits, messages, conversations and resolutions are not the same thing. To compare fairly, convert each quote to a cost per 1,000 conversations of the length you expect. For SiteTrellis, add the platform plan to the AI estimate from the calculator. For a managed product, divide the plan price by the conversations it really includes, and read how it counts them. Our comparison pages, starting with the alternatives hub, do this in the open, with last-checked dates.