Steam Recon

OSINT analysis of a Steam profile, targeted by a SteamID64, SteamID3 ([U:1:N]), SteamID2 (STEAM_X:Y:Z), a steamcommunity.com/profiles/… or /id/… URL, or a bare vanity name.

Needs a free Steam Web API key, configured under Settings → API Keys. Get one at steamcommunity.com/dev/apikey; the Steam account must not be limited (at least $5 spent). Steam allows roughly 100,000 requests per day per key.

Profile tab

A quick, single-request lookup: profile summary, ban record (VAC, game, community, trade), Steam level, game count, self-declared location, and deep-links to external lookups (steamid.io, SteamDB, csstats.gg, Leetify). A private profile still returns what Steam exposes publicly; fields it hides (game count, level, account creation date) show as unknown.

New Scan tab

Collects the target’s friends graph and analyzes it:

Close friends — ranks the target’s public friends by mutual-connection weight (closeness(i) = |friends(i) ∩ friends(target)|), not just friends-list order. The friend cap (default 200, up to 500) is applied to the oldest connections first, on the assumption an older friendship is more likely a genuinely close one; each kept candidate’s own friend list is then fetched to compute its mutual-connection weight against the target’s full friend set.
Geolocation hypothesis — a weighted vote of connected friends’ self-declared locations, aggregated hierarchically (country → region → city), with a high/medium/low confidence rating based on the leading candidate’s vote share, voter count, and location coverage. This is a hypothesis, not a fact, and is shown next to the target’s own self-declared location (if public) for comparison.
Friends graph — a radial diagram (self-drawn SVG, no charting library) of the target and its top close friends, sized and connected by mutual-connection weight.
CS2 cheater-probability report (optional, on by default) — a transparent, explainable heuristic across five signals: the target’s own VAC/game ban (weighted by recency), ban density among connected friends, low account investment (age/level/game count), CS2 stat outliers (headshot ratio, accuracy, K/D), and an accusatory-comment ratio from the target’s public profile comments. Every signal shows its own value, weight and contribution; a signal with no usable data contributes nothing rather than being scored as innocent. This is a heuristic estimate, not a verdict — see the in-app disclaimer.

A scan streams live progress and can be cancelled mid-run; past scans are kept under the History tab. Unlike Profile-tab lookups, a scan costs roughly one Steam Web API call per analyzed friend plus a handful more — the form shows an estimated call count before starting.

Deep-dive (algorithms, config values, confidence thresholds, live-verification notes): docs/architecture/steam-recon.md.