Know what support will cost. Answers from your docs. Team replies in Slack or Telegram.

A volatile support bill is hard on a small product

Micro-SaaS revenue can be lumpy. One launch, a new integration, or a migration can multiply customer questions in a week. A per-resolution bill makes the same growth event harder to forecast, while hiring a support person may not yet be realistic. The goal is not to make every question automatic; it is to answer the repeatable ones and keep the unusual cases visible to the founder.

Customers may ask “How many AI replies are included?”, “Does adding a teammate change the price?”, “What happens at the monthly limit?”, and “Can you cancel my account?” Publish the plan, limits, and self-serve steps clearly. A question that needs a private account change still belongs with a person, regardless of how small the product is.

• Track AI replies, not just conversation count.
• Budget for seasonal spikes and a human handoff path.
• Keep account actions outside the assistant’s claimed capabilities.

Flat pricing with a visible ceiling

Starter is $29/month for 1,000 AI replies, 3 seats, and 1 site. Pro is $99/month for 10,000 replies, 10 seats, and 5 sites. There is no charge per seat or per resolved conversation. The published plans let a founder check fit against actual usage instead of assuming unlimited AI.

The default estimate is 3 AI replies in one AI-handled conversation. That makes Starter’s 1,000 replies roughly 333 conversations. Fin would charge about $329.67 for that many resolved conversations at $0.99 each, before seats. The measures differ: a short exchange can use one MivoChat reply, while a longer thread uses more. Replies beyond the monthly limit are handed to a human rather than billed as an unplanned model call. Paid checkout is still coming soon; there is a no-card 14-day trial.

Make the product’s own help pages do the work

Connect the instructions customers need most: first login, data import, integrations, cancellation, and common error states. Use one canonical version of each answer and remove obsolete screenshots. MivoChat uses those connected pages to answer questions. If your docs contradict the current interface, the assistant cannot know which version a customer sees, so refresh the article before treating the answer as reliable.

An unclear case can go to a Slack thread or Telegram topic, where you or a teammate replies in context. The visitor stays in the widget. From there, the teammate can take over, hand the conversation back to AI, or close it. This gives a small product a manageable escalation path without requiring a separate helpdesk workflow for every question.

Add chat to the pages that explain the product

Install the snippet in your public site template, covering help and pricing pages where a customer is likely to ask for clarification. The standard script is: <script src="https://mivochat.app/widget.js" data-site="YOUR_SITE_KEY" async></script>. For Next.js, add the script in your layout with next/script. If you use another static site generator, add the snippet to the shared shell rather than copying it into every article.

Test a basic setup question, a price question, and an account action. Confirm the docs back the first two and the action goes to your team. Check what happens when the reply allowance is reached, and explain the human route in your public support policy. Review your plan usage monthly and after launches; a short period of higher traffic can reveal that the documentation needs another article or that Pro’s larger allowance fits better.

The assistant is not an account console

MivoChat cannot read a customer’s private data, see a subscription, change a plan, issue a refund, or perform cancellation. Say exactly what the public documentation establishes and route everything else.

A founder should own requests involving billing records, data deletion, access recovery, and bug investigation. The assistant should say it does not know when the pages do not answer the question. Keep the Slack or Telegram destination monitored enough that a handoff does not become a dead end. Predictable software cost helps only when the support promise itself is sustainable.