Steam Store Scraper — Games, Reviews, Prices & Charts

Scrape the Steam Store over plain HTTP — no browser, no residential proxies, no API keys. One Actor covering what used to need four separate scrapers: game search, full game details, user reviews, and live charts/feeds.

What it does

Find games

• searchQueries — keyword search across the Steam catalog ("roguelike deckbuilder", "cozy farming sim", ...)
• appIds / storeUrls — direct look-up of games you already know
• feeds — live Steam feeds: Top Sellers, New Releases, Coming Soon, Specials, plus Most Played (Steam Charts concurrent-players ranking, keyless)

Game details (includeGameDetails)

• Pricing: current price, original price, discount %, currency
• Review summaries: recent + all-time score, score description ("Overwhelmingly Positive"), total reviews, positive %
• Genres, user tags, categories, developers, publishers
• Release date, Metacritic score, DLC list, achievements count
• Platforms, supported languages, system requirements
• Header/capsule images, screenshots, trailers, recommendations

Reviews (includeReviews)

• Cursor-paginated — from a handful to tens of thousands per game
• Exact date windows (startDate / endDate, YYYY-MM-DD)
• Spreadsheet-friendly positive / negative verdict (no more TRUE/FALSE mangling in Excel)
• Playtime trio: hours at review, total hours, last-two-weeks hours
• Language, review type (positive/negative), purchase type filters
• Helpful/funny votes, weighted score, comment count
• Steam purchase / received-for-free / Early Access / refunded / Steam Deck flags
• Built-in sentiment analysis (score + positive/neutral/negative)

Monitor mode (newItemsOnly)

• Only outputs games never seen in previous runs (per search query or feed). Run on a schedule to track new releases, chart entries, or price drops.

Input

Output

Two record types in the dataset:

Game — appId, name, url, type, isFree, shortDescription,

price{currency, initial, final, initialFormatted, finalFormatted, discountPercent}

Review — reviewId, appId, gameName, recommendation (positive/negative), votedUp, reviewText, language, createdAt, updatedAt, author{}, playtimeForever, playtimeAtReview, playtimeLastTwoWeeks, votesUp, votesFunny, weightedVoteScore, commentCount, steamPurchase, receivedForFree, writtenDuringEarlyAccess, refunded, primarilySteamDeck, sentiment{score, label}, url, scrapedAt.

Pricing

• First 20 results per run are free (games + reviews combined).
• Then $2.50 per 1,000 games and $1.00 per 1,000 reviews (pay-per-event).

Notes

• Data comes from Steam's public Store endpoints; no login required.
• Review counts and scores match what the Steam Store shows.
• Release dates are ISO-8601 where Steam gives an exact date; vague dates ("Q4 2025", "Coming soon") are kept in releaseDateRaw.
• Sentiment analysis is a fast deterministic lexicon model tuned for English game reviews — great for triage, not a replacement for a trained classifier on nuanced text.
• Exact date windows on games with millions of reviews scan back from the newest review, so very deep historical windows take a while; recent windows stop early and are fast.

Changelog

• 0.2 — Feeds are now round-robin interleaved (quota shared fairly across all selected feeds instead of the first feed taking all); duplicates across queries/feeds no longer consume the maxGames quota (deeper results backfill); the sentiment key is omitted entirely when sentiment analysis is disabled.
• 0.1 — Initial release.