Steam Reviews Scraper

Pull the full review history of any Steam game as clean JSON: every review Steam still serves, newest first, with thumbs up or down, playtime at the time of writing, helpful votes, language, refund and Early Access flags, developer replies where they exist, and the game's all-time positive and negative totals. Filter by language, sentiment, date range and playtime before anything is fetched, so you only pay for rows you will read. $0.50 per 1,000 reviews, no start fee: a 30,000-review game costs about $15 once, and a daily watch on 10 games runs about $3 a month in incremental mode.

Built on Valve's public reviews data, no login, no API key required.

Why this one

• Full history, newest first, sliceable by date. sinceDate: "12 months" or a sinceDate/untilDate pair are applied by Steam itself, so a 1-million-review game narrows down to the two weeks after a patch without reading anything else.
• Steam-side filters, billed after filtering. Language, thumbs up or down, bought on Steam or activated with a key, minimum and maximum playtime, Steam Deck players, off-topic review bombs included or not. Filtered-out reviews are never fetched and never charged.
• Only new reviews on every scheduled run. Turn on incremental and the Actor remembers what it already returned per game; your dataset never needs deduping and you never pay twice for the same review.
• A summary row per game with the Steam review score bucket, all-time positive and negative counts, the count matching your filters, price, release date, Metacritic score and a last-N-months not-recommended share with the sample size next to it.

Who uses it

• Studios and publishers watch their own and competing titles after each patch: reviewType: "negative", sinceDate: "14 days", grouped by language.
• Community and support teams pull reviews from players with real playtime (playtimeMinHours: 20) and skip drive-by posts.
• Analysts and researchers get stable IDs, ISO dates, playtime and vote counts in one fixed schema across every language.
• AI agents call it through the Apify MCP server and read the field descriptions from the dataset schema.

Speed

Measured on the Apify platform: 100 reviews per request, about 1.5 seconds a page, so 3,000 reviews in under a minute and 100,000 reviews in about 25 minutes per game. Games run in parallel, pages within a game in sequence.

Use it from Claude, Cursor, or any MCP client

Add this URL as an MCP server in your client:

https://mcp.apify.com/?tools=actors,docs,kesar_chen/steam-reviews-scraper

For Claude Desktop, put this in claude_desktop_config.json:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com/?tools=actors,docs,kesar_chen/steam-reviews-scraper",
      "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" }
    }
  }
}

For Claude Code, one command:

$claude mcp add --transport http apify "https://mcp.apify.com/?tools=actors,docs,kesar_chen/steam-reviews-scraper" --header "Authorization: Bearer YOUR_APIFY_TOKEN"

Then ask things like:

• "Pull the negative reviews of app 413150 from the last 3 months and group the complaints."
• "Compare the not-recommended share of these three games over the last 12 months."
• "Every morning, fetch new reviews for my game and tell me if anything mentions crashes."

Review text is user-generated content: tell your agent to treat body as data, never as instructions.

Run it in the Apify Console

1. Paste app IDs, store URLs or search:game name into Games.
2. Optionally set Only reviews since, Review type, Languages or a playtime floor, and a maximum cost per run under Run options.
3. Click Start. The Reviews view holds one row per review, the Summary view one row per game. Export as JSON, CSV or Excel, or read it through the API.

Pricing

Pay per event, no subscription, no start fee.

Event	Price
Review scraped	$0.50 per 1,000 reviews
Game summary (one per game, with totals and score)	$0.002

Scenario	Reviews	Cost
Try it: one game, newest 500	500	about $0.25
One game, full history	30,000	about $15
10 games, daily incremental, ~20 new reviews each per day	~200 a day	about $3 a month

Apify platform usage is included. The monthly credit of Apify's free plan covers several thousand reviews. Paid Apify plans get a lower per-review price, shown on the Actor page.

Set a maximum budget per run in the Apify Console. When it is reached the run stops cleanly: reviews already written are kept and charged, nothing else is, and the summary row records how far it got. In incremental mode the saved state is only updated for games that completed, so a budget stop never loses data.

Input

Field	Type	Default	Notes
apps	list	required	App ID (413150), store URL, or search:name
languages	list	["all"]	Steam language codes: english, schinese, japanese, german ...
sort	select	recent	recent, updated, helpful, funny. recent is required for sinceDate and incremental mode
reviewType	select	all	positive, negative
purchaseType	select	all	steam (bought on Steam), non-steam (key activation)
sinceDate / untilDate	text	empty	2026-01-01 or 12 months, 30 days. Applied by Steam
playtimeMinHours / playtimeMaxHours	integer	0	Playtime at the time of writing. 0 = no filter
steamDeckOnly	boolean	false	Authors who played mostly on Steam Deck
includeOffTopic	boolean	false	Include reviews from periods Steam flagged as review bombs
keywords	list	[]	Keep only reviews containing one of these words
maxReviewsPerApp	integer	3000	0 = everything that matches
summaryWindowMonths	integer	12	Not-recommended share window in the summary row. 0 turns it off
incremental	boolean	false	Only new reviews on later runs, state kept per game
stateStoreName	text	steam-reviews-state	Named key-value store for incremental state. If a store with this name was created by another Actor, the run logs a warning and uses <name>-<actor id> instead
notifyWebhookUrl	text	empty	Slack, Discord, Feishu or any HTTPS webhook; see Alerts
notifyMaxRating	integer	1	Only reviews at or below this rating are posted
notifyMaxItems	integer	10	Reviews per message; the rest are counted
notifyEvenIfEmpty	boolean	false	Post a message even when nothing matched
includeAuthorId	boolean	false	Off by default; playtime figures are always included
searchResultsPerKeyword	integer	1	Games a search:name entry expands to

Minimal example:

{ "apps": ["413150"], "sinceDate": "12 months" }

Patch watch:

{ "apps": ["730", "1172470"], "reviewType": "negative", "sinceDate": "14 days", "languages": ["english"], "playtimeMinHours": 5 }

Output

One row per review plus one summary row per game. Use the type field or the two dataset views to tell them apart.

Review row:

{
  "type": "review",
  "reviewId": "237163238",
  "appId": 413150,
  "appName": "Stardew Valley",
  "language": "english",
  "recommended": false,
  "body": "Crashes on load since the last update.",
  "createdAt": "2026-10-20T12:34:56.000Z",
  "updatedAt": "2026-10-20T12:34:56.000Z",
  "votesUp": 12,
  "votesFunny": 0,
  "weightedVoteScore": 0.62,
  "commentCount": 1,
  "steamPurchase": true,
  "receivedForFree": false,
  "writtenDuringEarlyAccess": false,
  "primarilySteamDeck": false,
  "refunded": false,
  "playtimeAtReviewMinutes": 1830,
  "playtimeForeverMinutes": 4200,
  "playtimeLastTwoWeeksMinutes": 120,
  "authorReviewCount": 3,
  "authorSteamId": null,
  "developerResponse": null,
  "isNew": true,
  "scrapedAt": "2026-10-25T00:00:00.000Z"
}

Summary row:

{
  "type": "summary",
  "appId": 413150,
  "appName": "Stardew Valley",
  "appType": "game",
  "developers": ["ConcernedApe"],
  "publishers": ["ConcernedApe"],
  "releaseDate": "Feb 26, 2016",
  "isFree": false,
  "price": "$14.99",
  "metacriticScore": 89,
  "reviewScore": 9,
  "reviewScoreDesc": "Overwhelmingly Positive",
  "totalPositive": 1031073,
  "totalNegative": 15900,
  "totalReviews": 1046973,
  "totalMatching": 53808,
  "reviewsScraped": 3000,
  "coverageNote": "reached maxReviewsPerApp=3000",
  "window": { "months": 12, "from": "2025-10-25T00:00:00.000Z", "reviews": 3000, "negative": 41, "negativeShare": 1.4, "note": "sample 3000; coverage does not reach the whole window, counts are a lower bound" },
  "scrapedAt": "2026-10-25T00:00:00.000Z"
}

totalReviews is the all-time count; totalMatching is what matches this run's filters; reviewsScraped is what was written. negativeShare is only given when the window has at least 20 reviews. The window is always the last summaryWindowMonths, independent of sinceDate.

Limitations

• Steam has no star ratings and no review titles. recommended is the thumbs up or down; title is always null and exists only so that the schema lines up with the App Store Reviews Scraper.
• Developer replies are rare on Steam. developerResponse is filled when Steam provides one and null otherwise.
• Reviews from periods Steam flagged as off-topic review activity are hidden by default, matching the store page. Turn on includeOffTopic to include them.
• helpful and funny sorts do not page through the whole history reliably; use recent for full pulls.
• Steam may cache anonymous responses for a few minutes, so a review posted seconds ago can show up on the next run.

Alerts: post new low-rated reviews to Slack, Discord or Feishu

Fill notifyWebhookUrl with a Slack incoming webhook, a Discord channel webhook, a Feishu/Lark custom bot webhook, or any HTTPS endpoint, turn on incremental, and attach a Schedule. After each game the Actor posts the reviews written in that run that are at or below notifyMaxRating (default 1, meaning not-recommended reviews only), with the star rating, date, version where known, whether the developer already replied, and a link to the store page. Nothing is posted when nothing matched, unless notifyEvenIfEmpty is on. Unknown hosts receive a JSON body with the same fields, so Make, n8n, Zapier or your own service can take it from there. There is no extra charge for notifications.

An n8n template that does the same with room for translation or routing is in the templates folder.

Scheduling and integrations

Save your input as a Task, attach a Schedule and turn on incremental. Each run then returns only new reviews, which you can push to Slack, Google Sheets, Make, Zapier, n8n or any webhook through Apify integrations.

FAQ

Is it legal to scrape Steam reviews? This Actor reads public review data that Valve exposes without login or API key, and collects no private account data; the author's Steam ID is off by default. Collecting publicly available data is generally considered legal in the US and most of the EU. How you store and use the output, and whether it fits the terms of the services involved, is your responsibility. See Is web scraping legal?

Why not use Steam's own review export? Steam gives developers a dashboard for their own games only. This Actor works for any game, which is what you need for competitor analysis and market research.

How long does a run take? About 1.5 seconds per 100 reviews per game; games run in parallel.

What happens when my budget runs out mid-run? The run stops cleanly, keeps what was written, and the summary row records how far it got.

Is this affiliated with Valve? No. This is an independent tool and is not endorsed by or affiliated with Valve Corporation.

See also

App Store Reviews Scraper and Google Play Reviews Scraper: the same schema, filters and incremental mode for iOS and Android apps, so one pipeline can watch all three stores.

Hiring Signals Monitor: job opened / closed events and a hiring summary per company from public job boards, for watching studios' teams the same way you watch their reviews.

Support

Bugs or missing fields: open an issue on the Issues tab. Reports go to a human.