Customer support · Do this yourself

Let users file bug reports through an AI that dedupes and writes the ticket.

✅ Running in production; they pick the cheapest model that stays accurate enough.

No X data

Based on muzani on Hacker News (opens in new tab)

Who it’s for

Businesses that answer the same customer questions over and over.

Adapt it to your industry

Restaurant: hours, menu, allergens and bookings
Real estate: answer listing questions and book viewings
Clinic: opening hours, prices and how to book (never medical advice)

Step by step

Collect your 20 most common customer questions and your best answers in one document.
Open ChatGPT, Claude or Grok, paste the document and the prompt below, and ask it to improve and fill gaps.
Pick a chatbot builder (for example Chatbase, Intercom or Tidio) and upload that document as its knowledge.
Set a clear rule: if it isn’t sure, it hands over to a person and collects the customer’s contact details.
Test it with 20 tricky questions, including ones it should refuse.
Put it on your website or WhatsApp, and read every conversation for the first two weeks.

Ready-to-copy prompt

A user reported: "[text]". Search existing tickets for duplicates. If one matches, add this report as a comment. Otherwise write a ticket with title, steps to reproduce, expected vs actual, and severity, then submit it.

Start with AI

Opens the AI with this guide and the prompt filled in.

Open in ChatGPT (opens in new tab)Open in Claude (opens in new tab)Open in Grok (opens in new tab)

Cost: $0–20/month (free tiers or one AI plan)Our estimate
Time: 1–2 weekendsOur estimate

How to make money from it

Pricing idea: Set up support bots for businesses for a setup fee (for example $300–2,000) plus a monthly fee.

Finding your first 10 customers

Build a free demo bot using the public website of a business you want as a client, and send them the link.
Start with one industry so every demo is fast to make.
Ask happy clients for referrals to similar businesses.

Ideas, not promises. Results vary.

Details: source, numbers and tags

What they did

muzani built an AI tool where people report bugs or feature requests in plain language. It searches existing tickets to avoid duplicates, writes the ticket, and submits it. They praise Linear’s MCP as clean and well designed.

How it went

Running in production; they pick the cheapest model that stays accurate enough.

Worth knowing

Cost matters in production: pick the cheapest model that is still accurate.

Difficulty: IntermediateOur rating
Time to build: Not reportedOur estimate
Monthly cost: Not reported
Revenue: Not reportedWe only show revenue the source states
Engagement: 2 replies to HN commentComment in “Ask HN: Who is using MCP in production?” (198 points, 201 comments) · fetched Oct 15, 2026
X buzz: No X data

Posted Sep 4, 2026 · Added Oct 16, 2026

Original builder recipe

An LLM with tool calling
Linear MCP (or your tracker API)
A small intake UI

Collect the incoming channel (inbox, form, tickets) into one workflow.
Have the model classify and extract fields, returning structured JSON.
Route or draft replies; keep a human approval step for anything customer-facing.
Log results and review misclassifications weekly.