Eight steps, from idea to warranty.

Watch the AI turn a few sentences into a quote.

AI requirements and quoteExample

You

A booking mini program for our shops: customers pick a service and a time and pay online; staff manage shifts in an admin panel; a text message goes out before each appointment.

AI is confirming the requirements

  1. Where do customers book?

    WeChat mini programAppWeb
  2. How many shops?

    OneA chain
  3. How do they pay?

    WeChat PayIn store

Technical plan

  • Mini program and a web admin panel
  • TypeScript API service with PostgreSQL
  • Deployed on your cloud server

Tasks

  1. Architecture and API designFable 5.1max
  2. Database and permissionsOpus 5.5max
  3. Booking and shiftsOpus 5.5max
  4. WeChat Pay and refundsOpus 5.5max
  5. Mini program front endOpus 5.5max
  6. Admin panelOpus 5.5max
  7. Text remindersOpus 5.5max
  8. Acceptance tests and fixesOpus 5.5max
TaskModelTokensCost
Architecture and API designFable 5.1max20M¥524
Database and permissionsOpus 5.5max30M¥332
Booking and shiftsOpus 5.5max60M¥663
WeChat Pay and refundsOpus 5.5max40M¥442
Mini program front endOpus 5.5max80M¥884
Admin panelOpus 5.5max70M¥774
Text remindersOpus 5.5max15M¥166
Acceptance tests and fixesOpus 5.5max50M¥553
AI usage¥4,338
Overrun budget, 20%¥868
Labor, 8 days × ¥1,800¥14,400
Tax (VAT 6%)¥1,176
Total¥20,782

Day 1

  • Contract signed
  • Deposit received
  • Schedule set

Claude Code at workOpus 5.5max

  1. Architecture and API design
  2. Database and permissions
  3. Booking and shifts
  4. WeChat Pay and refunds
  5. Mini program front end
  6. Admin panel
  7. Text reminders
  8. Acceptance tests and fixes

Accepted

  • Every test passes
  • You sign off item by item
  • Balance received

Delivered

  • Source code and commit history
  • Deployed on your server
  • User and operations guides
  • Accounts and keys handed over

3 months

  • Problems fixed free of charge
Day 90 of 90
An example quote; yours follows your requirements. At US$1 = ¥6.7, with a typical token mix: 85% cache hits, 8% cache writes, 2% input, 5% output.

How the quote is built

Four lines of cost, one itemized quote.

  1. Requirements, settled first

    We read your documents and description, confirm each point with multiple-choice questions, settle the plan, the stack and the architecture, then split it all into buildable tasks.

  2. AI usage, by the token

    Each task’s token use is estimated for the model it runs on and priced at Claude’s official rates, with 20% added as an overrun budget.

  3. Labor, ¥1,800 per day

    Covers communication, waiting, decisions and delivery.

  4. Tax

    Calculated by law, according to the invoice type.

Claude’s official rates

Per 10M tokens

ModelInputCache writeCache hitOutput
Claude Fable 5.1Architecturethe top model, max effort$100$125$2.50$500
Claude Opus 5.5Development and acceptancethe next model down, max effort$40$50$2$200

In US dollars, from Claude’s official price list (November 2026); cache writes at the 5-minute rate. Converted to yuan at the day’s rate when we quote. See the official price list

Guangzhou at dusk from the office window

Send us your requirements.

A document or a few sentences, either works.

Partnerships
[email protected]
Careers
Join us