Home Marketplace Build a Price Monitor with Claude Code
Build a Price Monitor
with Claude Code
Finished code teaches you nothing — you didn't write it, so you can't extend it, and the day it breaks you're stuck. This is a build-along instead: a guide containing every prompt, in order, that builds a working competitor price monitor — watchlist, robots-aware fetching, CSS-selector extraction with a currency fallback, append-only price history, and change alerts to console or any webhook. Plus the finished, live-tested repo to check yourself against. Polite by design: it reads robots.txt before it crawls, rate-limits itself, and never scrapes behind logins.
Instant download · Commercial licence · Free updates for life · 7-day refund
The problem
"Clone this repo"
teaches you nothing.
Every scraping tutorial ends the same way: here's the finished code, clone it, run it. It works — until the site redesigns, the selector dies, or you need it to watch something the author didn't. Then you're maintaining code you never understood, and the fastest fix is abandoning it.
A build-along inverts that. You build the monitor yourself by prompting Claude Code through it — the guide contains every prompt, in the order that produced the working repo — so by the end you understand each piece well enough to extend it. The finished repo is included, but as your answer key, not your starting point. And the pattern is the real product: the same skeleton rebuilds in an evening for job postings, stock levels, review counts, or any public page that changes.
Repo — answer key, not crutch
What's inside
The guide, the repo,
and every prompt.
Chapter 0 — the rules of the road
Ethics first, because scraping discipline is part of the lesson: public pages only, robots.txt respected, polite request rates — and the monitor refuses to do otherwise by design. It identifies itself honestly and honours crawl-delay. Most "my scraper got blocked" stories start with skipping this chapter.
Never — behind logins
Chapters 1–2 — the contract prompt, then polite fetching
The build starts with the contract-first prompting pattern: you tell Claude Code what the program must and must not do before any code exists, so every later prompt refines instead of rescues. Then fetching — rate-limited, robots-aware, with honest headers.
Chapter 3 — extraction that survives redesigns
CSS-selector extraction per watched page, with a currency-pattern fallback for when the selector dies — the monitor degrades gracefully instead of silently reporting nothing. This chapter is where most homemade scrapers fail; here it's designed for from the start.
Fallback — currency regex
Chapters 4–5 — history, change detection, alerts
Append-only price history per watched item, change detection against the last recorded price, and alerts to console or any webhook — Slack and Discord ready. Chapter 5 ends with scheduling: one cron line and it checks daily without you.
Alerts — console / webhook
Chapter 6 + appendices — making it yours
Swapping in your competitors' pages, adapting the skeleton to other monitoring jobs, Appendix A: the full prompt log from the real build, and Appendix B: a verification checklist so you can prove your build matches the tested one.
Appendix B — verify your build
The finished repo — live-tested
monitor.py, watchlist.json, requirements.txt and README — runs out of the box against a legal practice site, with fetch → extract → history → change detection all verified before shipping. Guide ships as PDF and Markdown. Requires Python 3.9+ and Claude Code.
Guide — PDF + Markdown
Who it's for
People who'd rather build it
than clone it.
Developers learning AI-assisted building
You want to learn how a real project gets built by prompting — contract first, piece by piece, verified at each step — not just read the final code.
Operators watching competitor prices
You check competitor pricing manually, rarely, and too late. The finished repo plus one cron line makes it daily and automatic.
Anyone who needs the pattern
Prices today — job postings, stock levels or review counts tomorrow. The same skeleton rebuilds in an evening for any public page that changes.
Not a fit if you want a hosted price-tracking service. This is a build-along — you run the code on your own machine and learn the pattern. If you won't open a terminal or install Python, this deliberately isn't for you.
Honestly
What this can't do.
It will not scrape behind logins — ever. The monitor watches public pages only, respects robots.txt, and rate-limits itself. If a site's robots.txt disallows crawling, the tool refuses, by design. That's not a limitation to work around; it's the lesson.
Selectors are per-site work. The guide teaches you to find the right CSS selector for each page you watch, and the currency fallback catches redesigns — but pointing the monitor at a new competitor takes a few minutes of your attention. It is not magic autodetection.
You need Claude Code to do the build-along. The finished repo runs on plain Python 3.9+, but the point of the product is building it yourself by prompting. Anthropic bills your Claude usage; there are no fees to me.
Refuses — robots.txt disallows
Questions
Before you buy.
Is scraping competitor prices legal?
Watching public pages politely is how every price-comparison site and search engine works, but rules vary by jurisdiction and by site terms. This monitor is built to stay on the defensible side of the line: public pages only, robots.txt respected, honest identification, rate-limited requests, and never anything behind a login. Chapter 0 covers the ethics and the reasoning before any code is written. If a site tells crawlers to stay away, the tool stays away.
Why did my last scraper get blocked, and why won't this one be?
Most blocked scrapers earned it: they ignored robots.txt, hammered the server with rapid requests, and lied about who they were. This monitor does the opposite — it fetches robots.txt first and honours crawl-delay, waits politely between requests, and identifies itself honestly. No scraper is unblockable, but polite scrapers watching a handful of pages daily rarely have problems, because they look like what they are.
What exactly is a "build-along"? Do I get finished code or not?
Both, in the right order. The guide walks you through building the monitor yourself in Claude Code — every prompt from the real build is included, in sequence, with what each one produced. The finished, live-tested repo is also in the download, so you can run a working version in five minutes and use it as the answer key for your own build. The learning is the build; the repo is the safety net.
What happens when a competitor redesigns their site?
The primary CSS selector for that page stops matching — and instead of silently reporting nothing, the monitor falls back to a currency-pattern search so you usually still get a price, and you can see the selector needs updating. Fixing it is editing one line in watchlist.json. Chapter 3 teaches you to pick selectors that survive most redesigns in the first place.
Can I adapt this to monitor things other than prices?
Yes — that's the point of buying the pattern rather than a price-tracking subscription. The skeleton is fetch politely → extract a value → append to history → alert on change. Job postings, stock levels, review counts, ranking positions: Chapter 6 covers the adaptation, and the same contract-first prompting rebuilds it for a new target in an evening.
Do I need to be a Python expert?
No. You need Python 3.9+ installed, Claude Code, and the willingness to read what the prompts produce. The build-along assumes you can run commands in a terminal; it does not assume you could write monitor.py unaided — that's what you're learning. Commercial licence, and if it isn't useful, reply to your receipt within seven days for a full refund — you keep the files.
Know their price before your customers do.
Instant download. Commercial licence. Free updates for life. Seven-day refund, no justification needed.
Fetch → extract → history → alerts, all verified live before listing
Where this came from
We build scrapers and monitors for our own lead pipeline, and the polite-by-design rules here are the ones our production crawlers follow. Read why your scraper got blocked — and the polite pattern that doesn't, or if you'd rather have monitoring built for you, that's the day job.