A long list of checks on a plugin page doesn't guarantee much. What matters is whether those checks work on your version, what actually triggers a ban, and what happens to TPS at full player count. Ask these questions first, then test the answers on your own server.
| Question | A good answer |
|---|---|
| 1. Software and version | Your version is proven on production servers; ViaVersion clients are checked the same way |
| 2. Fronts | Combat, movement, building, inventory, packets and automation — not just combat |
| 3. How it decides | Bans on evidence piled up across separate episodes, with a report |
| 4. Setbacks | Silent mode: the cheater can't see which feature got caught |
| 5. Load | Checks run off the main thread, and you can measure the load |
| 6. Updates | New bypasses get closed fast; support answers directly |
| 7. Price | A model that makes sense for your player count; a free plan without cut-down protection |
| 8. Honesty | Every claim can be checked against code or docs |
1. Does it run on your server software and version?
Start with compatibility: an anticheat that doesn't support your server software or version is out straight away. Look past the headline to where it's actually been proven — "supported" and "running on production servers" aren't the same thing. If you haven't picked your server software yet, read which server software to choose first.
Ask about players too: if they join from different versions through ViaVersion, the anticheat should check all of them the same way. On a proxy network, find out where it goes — on the proxy or on each game server.
DGuard: Paper and Leaf 1.21–26.3. The main version is 1.21.11: DGuard is tested on it and runs on it on production servers; the others are supported but haven't been run on production servers yet. 1.21–26.x clients joining via ViaVersion are checked the same way, and on a BungeeCord or Velocity network DGuard goes on each game server with its own key. Folia isn't supported, and old versions like 1.8 or 1.12 aren't on the list. The full list is on the versions page.
2. Does it cover every front, not just combat?
Cheaters rarely stick to one cheat: KillAura in fights, Speed for movement, Scaffold for building, Nuker in the mines. An anticheat that's only strong in combat leaves a survival or anarchy economy wide open, and one that's only strong at movement misses combat automation.
Check all five fronts: combat, movement, building and world, inventory, and packets and automation. Anticheats differ most on inventory and automation — here's one kind of cheat from each front:
| Cheat | DGuard | GrimAC | Vulcan | Polar | Intave | Matrix |
|---|---|---|---|---|---|---|
| KillAurahits targets by itself | partly: no dedicated check — only indirectly, via Reach and packet order | |||||
| Speedfaster than the game allows | ||||||
| Nukerbreaks everything around at once | partly: FastBreak limits the pace; the dedicated MultiBreak is experimental | partly: not named — only a generic block-interaction check | partly: not named — only the break time of a single block | partly: not named — only the generic block-breaking checks | ||
| AutoArmorputs armor on by itself | partly: no dedicated check — only inventory clicks while moving | partly: no dedicated check — only inventory click speed | partly: not named — only a generic inventory check | partly: no dedicated check — only inventory click speed | partly: not named — only how fast inventory actions are | |
| AutoFishfishes by itself |
DGuard: a check for each of the 47 kinds of cheats in the comparison, with 11 checks and 80+ rules watching each player at once. What every rival covers is broken down in The best Minecraft anticheat.
3. How does it decide who gets banned?
Lag, high ping and allowed mods sometimes produce odd packets. An anticheat that bans on a single flag will sooner or later ban an honest player — and that means appeals, arguments with staff and a player who doesn't come back. Ask what exactly triggers a ban and whether you can see why afterwards.
A good answer is evidence collected across separate episodes, with a report for every ban. Then a disputed ban gets reviewed on facts, not on someone's word. More on that in false bans.
DGuard: every flag is evidence weighted VL × the rule's weight, and it halves every day. A ban comes only at 400 points across three separate episodes. Allowed mods and keybinds are listed in the code and produce no evidence. Every ban comes with a report breaking down the evidence, and if a ban turns out wrong, support goes through the fight tick by tick and fixes the rule.
4. Setbacks or silent mode?
A setback moves the cheater back to their previous position; a cancelled hit stops them from landing it. The upside: the cheat's effect is stopped right away. The downside: the cheater sees which feature got caught and tweaks the cheat's settings until they find a bypass.
Silent mode cancels nothing: the anticheat quietly gathers evidence and punishes once there's enough. The cheater doesn't know they've been caught, so there's nothing to tune a bypass against. Find out what the anticheat does by default: Vulcan, for example, only watches by default, with setbacks available per check; Polar doesn't punish movement or reach and mitigates the action instead.
DGuard: runs silently — no cancelled hits and no setbacks. Strictness is yours: ban or kick per cheat feature, with evidence weights and thresholds in the config, from lenient to zero tolerance.
5. How much main-thread time does it take?
A Minecraft server runs the game at 20 ticks per second, 50 ms per tick. Anything an anticheat does on the main thread is taken away from the world, mobs and players, and at high player counts it shows in TPS. Ask where the checks run, and measure it yourself — anticheat performance shows how.
DGuard: checks run on network threads, not the main thread, so TPS stays at 20. On a live server that's 0.03% of one core on the main thread: 0.38 ms of a 10.9 ms tick with 121 players. Run /dguard perf to see the load on your own server.
6. How fast are new bypasses fixed?
Cheats get updated all the time, and an anticheat that stands still falls behind. Ask how often new versions come out, whether they install themselves, where to report a new bypass and who answers — a developer, a bot or nobody. A public changelog is a good sign.
DGuard: a new bypass gets closed within a couple of days, new versions install themselves when the server restarts, and every release is listed in the changelog. DGuard's combat model was trained on real cheaters on a live server with 500+ online. Support answers you directly — in the panel, through the Telegram bot or via the form on the site, even without an account.
7. What exactly are you paying for?
Anticheats price in different ways, and the same amount means different things for a server with 20 players and one with 500. Do the maths for your player count, and check what the free plan includes, if there is one.
| Model | Example | What to check |
|---|---|---|
| Free, open source | GrimAC | What's covered now versus promised: GrimAC's combat heuristics are planned for a paid version |
| A license per server | Vulcan — $19.99 per server | Every new server in the network needs another license |
| Pay per player-hours | Intave — €0.79 per 1,000 player-hours | What the free plan includes: Intave's aura and scaffold detection is paid-only |
| Subscription by player count and servers | Polar, DGuard | Whether there's a free plan, and what happens if you outgrow the plan |
DGuard: free up to 30 players online — no time limit, no card. Protection is the same on every plan; you pay for players online and the number of servers, weekly, monthly or for 3 months at a discount. If you outgrow your plan, nothing switches off and players can still join. See pricing for details.
8. Can you verify its claims?
"Catches every cheat" isn't an argument. An honest anticheat says what it catches, and the best ones also say what they don't — Vulcan, for example, lists that right on its page. Look for primary sources: the source code, the docs, the config with its checks.
Don't trust promises that a plugin catches X-Ray, ESP or Freecam: those cheats only change what the client draws, and the server can't see them directly. X-Ray is handled by Paper's anti-xray.
DGuard: every mark for GrimAC, Vulcan, Polar, Intave and Matrix in the comparison links to their code or docs, and "partly" doesn't count. You can look around the panel, with its log and reports, in the demo without signing up.
Test the answers on your own server
- 1
Run the new anticheat alongside
Don't remove your current one. Add the new one on lenient settings — DGuard is free up to 30 players online, and on the beta servers it was tested exactly like this, next to other anticheats.
- 2
Watch for a week
Compare the logs: who both caught, who only the new one caught, who only the old one caught. Read the reports, not just the counters.
- 3
Check your honest players
If players you know to be legit show up in the log, go through their reports. Better to find a false flag before bans are switched on.
- 4
Move the punishments over
Turn on punishments in the new anticheat, and only then remove the old one. Step by step: How to switch anticheats.
FAQ
Which anticheat should I use for my Minecraft server?
One that runs on your server software and version, covers every front, bans on evidence and stays off the main thread. In our comparison only DGuard has a check for all 47 kinds of cheats, and its AI Combat neural network reads combat on top.
Do I need an anticheat if I have moderators?
Yes: a moderator can't watch every player at once and can't see packets. An anticheat gathers evidence around the clock, and moderators review the disputed cases from its reports.
Is a paid anticheat better than a free one?
Not necessarily. The free GrimAC is strong at movement but has no dedicated KillAura or AutoClicker checks, and Intave's free plan doesn't catch aura or scaffold. Compare the checks, not the price.
Can I test an anticheat before paying?
Yes: run it next to your current one and compare the logs after a week. DGuard is free up to 30 players online with no time limit, and you can look around the panel in the demo without signing up.