Skip to content
agentusehub

Fill a supermarket cart from regular purchases and sales

A developer built a Claude Code skill using Silpo's MCP server to re-add usual groceries, check sales and fill the real cart within a budget, leaving checkout to the human.

Time saved
Not reported
Setup time
Not reported
Running cost
Not reported
Cost range
Not reported

Use case overview

A developer built a Claude Code skill using Silpo's MCP server to re-add usual groceries, check sales and fill the real cart within a budget, leaving checkout to the human.

Reported experience

  1. What they did

    The developer wrote silpo-agent-cli, a terminal CLI and Claude Code skill that talks to Silpo's official MCP server. You give Claude a plain-language instruction, such as reordering usual groceries, swapping in a sale brand and staying under 1500 UAH. Claude reads purchase history, weighs costs, checks stock and fills the real cart. It stops at a ready cart, and the human handles checkout, delivery address and payment.

  2. How it went

    The source gives no measured results. It says weekly shopping is about 80% repeat purchases, and re-adding them by hand takes 15–20 minutes. It includes a demo GIF and video but no test data.

  3. Worth knowing

    Checkout, address and payment are deliberately left to the user, so the cart still needs your review. Install with pip or run it through uvx.

Example prompts

Try with Claude Code
Create a skill that connects to [supermarket]'s MCP server, re-adds my usual groceries from [list or past orders], checks current sales for them, and fills the real cart within a budget of [amount]. Suggest cheaper swaps where there are sales but ask me before swapping. Stop when the cart is filled and show me the cart and total; do not check out.
Coding · Advanced
  1. Write migration plan
  2. Agent edits in batches
  3. Tests per batch

Ship a 40-file refactor with a coding agent

Plan, execute and review a cross-cutting refactor in small batches, with Claude Code making the edits and GitHub Copilot reviewing each PR.

Claude Code · GitHub Copilot
Coding · Advanced
  1. Run coverage
  2. Pick 3 riskiest files
  3. Write + run tests

Generate missing unit tests for untested modules

Claude Code reads your coverage report, picks the riskiest untested modules and writes unit tests that pass, one small PR at a time.

Claude Code · GitHub Copilot · Slack