PriceRunner.dk

scripts/pricerunner.py wraps PriceRunner's unofficial frontend JSON API (no auth, stdlib Python). Every command prints one line of compact JSON; pipe through jq to trim. Prices are DKK. The API is reverse-engineered from the site's JS and can break without notice — if a command starts returning HTTP 404/400, say so rather than guessing.

PR="python3 <skill-dir>/scripts/pricerunner.py"   # substitute the absolute path of this skill directory
$PR <command> -h                                   # full flags per command

Commands

Command	Use
search "<q>"	Free-text search. Returns products, the filters PriceRunner applied, and facets (filter ids + option ids) to refine with.
categories [REGEX]	Category tree. Find the cl id for a product type.
filters CAT	Every filter id available in a category (PRICE, BRAND, RATING, numeric spec ids…) with type.
facets CAT ID…	Option ids + product counts for those filters; ranges get min/max buckets. Pass -f to see counts under current filters.
list CAT	Browse a category with filters and sorting. Also returns subcategories.
offers PRODUCT	Shop offers, default Danish in-stock new, cheapest incl. shipping first.
history PRODUCT	Daily lowest price: current, all-time-in-window low/high/median, currentPercentile (0 = cheapest ever seen), downsampled series.
info PRODUCT	Lowest national/international price, review score, pro reviews with pros/cons.
compare PRODUCT…	Spec table; by default only rows that differ.
similar PRODUCT	Alternatives in the same category.
deals	Current price drops.

PRODUCT accepts a bare id (3216399795), 94-3216399795, or a pricerunner.dk /pl/ URL.

Filters

-f takes ID=VALUE and is repeatable; the af_ prefix is added for you.

• Options (OR within one filter): -f BRAND=162,1766
• Range: -f PRICE=500_1500, -f PRICE=_1500 (max only), -f PRICE=2000_ (min only). Same for numeric spec filters (e.g. battery hours).
• Rating: -f RATING=4_
• In stock only: -f ONLY_IN_STOCK=true (works in list; search ignores it)

Filter ids are numeric for specs (e.g. 59671607 = IP-klasse in headphones) and per category — always look them up with filters then facets; never guess option ids.

Category scoping differs by command: search -c 94 or search -c 100003567-100014541 (sub id); list 94 --sub 100003567-100014541.

Workflow

1. Vague request ("cheap noise-cancelling earbuds"): search first. Its appliedParams shows what PriceRunner auto-extracted from the query (often an af_CATEGORY); pass --no-extract if that narrowed wrongly. Use categoryId from the results for the next steps.
2. Spec-driven request ("laptop with 32 GB RAM under 10.000 kr"): categories 'bærbar' → filters 27 → facets 27 <ramId> -g 32 → list 27 -f <ramId>=<optId> -f PRICE=_10000 -s PRICE_ASC.
3. "Where do I buy X cheapest?": offers. Quote total (price + shipping), stock and delivery days; mention --international if summary.internationalOffer.minPrice is notably lower.
4. "Is this a good price / should I wait?": history -i ONE_YEAR (or TWO_YEARS). Compare current with lowest and median; a low currentPercentile means it is near its cheapest. Note seasonal dips (Black Friday, November) visible in series.
5. Cheapest-first listings (-s PRICE_ASC) surface miscategorised junk and one-shop marketplace listings. Set a realistic PRICE minimum and prefer products with several merchants, or sort by POPULARITY/RATING and filter by price instead.
6. Choosing between models: compare for differing specs, info for review scores and pro pros/cons.

Present a short answer with product names, prices and pricerunner.dk links — not raw JSON. Offer urls go through Klarna's clickout redirect; prefer linking the product page url from search/list/info.