UNDER THE HOOD
Everything itwatches for.
The full catalogue, the scoring model behind every ban, and the APIs you can build on. Nothing here is marketing — it is read straight from the shipped product.
Detection catalogue
Nothing vague. 88 named checks, and counting.
This is the real catalogue the resource ships with — the same names, descriptions, severities and default actions you will see in the configuration panel. Every one of them can be switched off, re-thresholded or re-pointed at a different action, per server.
Detections
14 categories
Ban by default
the rest kick, flag or log
Hold per-player state
the others are stateless
Flagged, never auto-banned
high false-positive risk
Showing 24 of 88
−100 Integrity on a confirmed hit
−40 Integrity on a confirmed hit
−18 Integrity on a confirmed hit
−8 Integrity on a confirmed hit
−3 Integrity on a confirmed hit
Integrity Score
Ban the pattern, not the moment.
Every player carries a 0–100 behavioural score. One ambiguous hit costs a few points; a cheat that keeps tripping the same check converges on a ban on its own. It is what lets ARGUS act on evidence that no single detection could justify.
Auto-banning on score alone is off by default. You decide what happens here.
Entity creation and event rates are tightened for this player.
Every action the player takes is recorded in full detail.
The resting state of a healthy player.
Nothing is tightened. The player is invisible to the system.
Everyone starts at 100
A new connection is trusted until it earns otherwise. Nobody has to prove anything to play.
Repeat hits escalate
The same detection tripped again inside a session adds 35% of its penalty each time, and the multiplier stops at 3.1× after six repeats. A cheat that trips one check fifty times converges on a ban; a single noisy hit does not.
Risk-graded penalties
A detection graded medium false-positive risk costs 75% of its penalty; high risk costs 40%. Uncertainty is priced in rather than argued about afterwards.
Recovers at 2 points an hour
Clean play returns points at a rate you set, and the score never climbs past 100. A bad night is not a permanent record.
Watch below 60
Full-detail recording switches on so that when the next hit lands, the evidence is already there.
Restrict below 35
Spawn and event rate limits tighten for that player alone — the rest of the server is untouched.
What a confirmed hit costs
Before escalation and risk grading.
Resource API
Teach it what your scripts are allowed to do.
Most false positives are a script doing something legitimate that looks exactly like a cheat. Rather than blinding the anti-cheat, tell it what is about to happen — one line, server-side Lua, no framework assumptions.
Authorise, do not ignore
Before your script does something a cheater could also do, say so. The grant expires on its own, so the hole closes even if your script crashes mid-way.
Route your own logic through the pipeline
reportPlayer hands your anti-exploit checks the same thresholds, Integrity Score, panel history, Discord alerts and ban decision the built-in detections get.
Protect your events
Mark any of your server events as server-to-server only. A client that triggers one is flagged instantly, and block also cancels the call.
Everything the panel can do, over HTTPS
Bans, players, detections and live commands are all available with an API key you create in the panel. Webhooks push ban.created, ban.removed and detection.raised the other way, HMAC-signed.
-- duration is seconds; -1 (the default) is permanent.
local banId = exports['argus']:banPlayer(source, 'Duping money', { item = 'cash' }, -1)
exports['argus']:kickPlayer(source, 'AFK in a restricted zone')
-- Lift a ban by the public code the player was shown.
exports['argus']:unbanPlayer('AR-7K3QPX', 'appealed and accepted', 'AdminName')
local exists, ban = exports['argus']:getBanInfo('AR-7K3QPX')