The best Discord anti-raid and anti-nuke bots

A raid is a flood of accounts joining at once. A nuke is one trusted person deleting every channel. They need different defences, and most bots sell only one. Of the sixteen moderation and security bots we checked, three ship real anti-nuke: Wick, Security Bot and ProBot. Wick covers anti-nuke and is the only one of the sixteen that attempts recovery afterwards.

The distinction that decides which bot you need

Three separate problems get filed under “raid protection”, and buying the wrong one is the most common mistake in this category.

1. Join-rate raid detection. Fifty accounts join in ninety seconds. The bot notices the rate and locks the door. Discord’s own tools do nothing here at all.
2. Per-account screening. One account joins and something about it is wrong — days old, no avatar, a username matching a pattern. This is a screen, not an alarm, and it runs on every join.
3. Anti-nuke. Someone with permissions mass-deletes channels, mass-bans members or hands dangerous permissions to a role. The attacker is inside, and rate-limiting joins does nothing.

Discord’s native AutoMod covers none of the three. It filters message content — five trigger types, a maximum of six custom keyword rules, three preset wordsets and four actions — and it has no join-rate protection of any kind, no kick, no ban and no case record. The entire anti-raid market lives in that gap. It is also powerless against an administrator: it inspects messages, and it does not restrain the person who can delete the channel the message was in.

How this ranking was built, and what it is not weighted by

Every capability and price below was read from the vendor’s own site or documentation on 30 August 2026. The ranking weighs four things:

• Whether the defence is free, because raid protection you have not bought is protection you do not have on the night it matters.
• Whether it covers joins, behaviour and nukes, or only one of them.
• Whether the protection defends its own configuration, since an attacker with dashboard access can otherwise switch it off.
• Whether anything survives the incident — a case record, a log, a restore.

It is not weighted by install count, by directory position, or by how loudly a vendor markets itself as a security product. Several of the bots below are far more popular than several others, and popularity is covered separately in the most popular Discord bots.

The comparison

Beemo — free raid detection, and the most opinionated

Beemo does one thing and gives it away permanently. Its raid detection is free forever, cannot be configured, and cannot be disabled. That reads like a limitation until you read the vendor’s reasoning: “The algorithm is not configurable because it is designed to work dynamically… no other configuration required (or available).”

The argument holds up. A per-server rule engine only knows what one server has seen. Beemo’s /report-raid loop feeds detections back into a shared model, so every protected server improves detection for every other one. That is a network effect no configurable rule engine can have, and it is the strongest structural argument in this whole category.

What it does not do: behavioural protection, anti-nuke, or anything about an insider. Its moderation logs are the paid part, at $4.99/month or $53.89/year. Beemo is a layer, not a security suite, and it is best run alongside something else.

Verdict: install it. It costs nothing and it is better at its one job than the general-purpose alternatives.

Wick — anti-nuke as the whole product, and the only repair attempt

Wick is a pure security bot: an anti-nuke monitor, a heat-based automod, a join gate and join raid pair, and a verification gate. Nothing else — no leveling, no tickets, no economy.

Two things put it first on nuke defence.

Its protection defends itself. Bypassing quarantine, adding dangerous permissions to any role, and changing the vanity URL all quarantine the actor. Most protection products guard the server and leave their own configuration surface open, which is exactly where an insider goes first.

Of the sixteen moderation and security bots we checked, it is the only one that attempts recovery. Panic Mode deploys an isolated process to reconstruct pre-nuke state, then restores a structure-only backup image. Every other product in this category detects and punishes; Wick tries to put the server back.

Wick also has a Discord AutoMod integration that writes rules directly. Since v5.3.0 its dashboard creates, edits and removes Discord’s own AutoMod rules and attaches Wick’s punishments to them. That is the correct architecture, because Discord blocks a message before publication and no bot can do that. Any article telling you third-party bots replace native AutoMod is wrong; the right answer is to drive one from the other.

The caveat that matters, and it is not small. Wick’s anti-nuke explicitly exempts anyone holding a role above Wick’s own. So does Security Bot’s, at or above its own role. An insider who can move roles defeats both, and neither records that person editing the protection’s settings.

Pricing: $5/month per server, or $20/month covering 12 servers — about $1.67 per server, materially cheaper for anyone running several communities. The prices only render after clicking the plans control on the page; a plain fetch returns nothing.

Verdict: the reference product for anti-nuke. If a rogue administrator is your threat model, this is the one to start from.

Security Bot — free where everyone else charges, paid where everyone else gives

Security Bot inverts the category. Anti-nuke and anti-raid are free; chat automod is premium. Everyone else does the opposite. A free Security Bot server is protected from its own administrators and not from spam, which is a genuinely defensible position and also a surprising one — assume the usual shape and you will misjudge it.

One thing to check before buying: the pricing page shows $3.49/month in bundles of 1, 3 and 7 servers, plus $11.99/month Premium+, while the documentation and homepage narrative say $3.50 and $8.50. The two surfaces disagree and we could not resolve which is current. Read the checkout before you commit.

Verdict: the cheapest real anti-nuke, and the right pick if your risk is insider damage rather than message spam.

ProBot — anti-nuke included, anti-raid behind the paywall

ProBot ships a real anti-nuke module, which puts it in a group of three. Its anti-raid is paid, which puts it in a group of two alongside Circle. Tier 1 is $5/month or $40/year; Tier 2 is $10/month or $80/year and its genuine additions are a custom bot and ownership transfer, not protection.

Worth correcting a claim several comparison articles repeat: ProBot’s Advanced Protection appears on both tiers on its own pricing page. Only the Tier 2 description mentions anti-raid, which is the likely source of the confusion.

Verdict: reasonable if you already run ProBot for its free leveling. Not the reason to install it.

Sapphire, Gaius, GiselleBot and Zeppelin

Sapphire is generous and has a real gap: no join-rate anti-raid and no anti-nuke, checked against its own documentation. What it does have is seven free Join Guard filters, ANDed together, plus 12 automod modules and 80+ log types free. Excellent moderation bot, wrong tool for a raid.

Gaius gives away a heat and strike engine, account-age screening and voice moderation, and its raid protection is partly free. It has no anti-nuke. It also declares its own documentation “SEVERAL YEARS OUT OF DATE”, in capitals, and publishes no price anywhere — Gaius Plus sits behind Discord OAuth on a beta dashboard.

GiselleBot has free behavioural protection with an explicit Ban > Mute > Kick > Warn precedence and 10 triggers including QR-code detection inside images. Its paywall is unusual: all logging is paid, across User, Server and Voice loggers. It has no anti-nuke, and none of its four Patreon tiers is priced on any reachable page.

Zeppelin is a moderation engine built around configuration depth — 34 automod triggers, 19 actions, 66 log types — and it is the one most readers cannot use. Its own documentation says hosted access is “granted on a case by case basis”, the whitelist is enforced in its source by a plugin that leaves non-whitelisted servers, and there is no invite link on the site. Self-hosting is Docker on Linux under the Elastic License. It has no anti-nuke as a shipped product, though one is constructible from automod triggers feeding a counter.

Where Vetox fits

Vetox splits the problem the same way this article does, and prices the two halves differently.

Protection is free on the Starter tier, and it is broad: 20 distinct rule types covering mass channel and role deletion, permission changes, mass bans and kicks, webhook and emoji abuse and the rest, with every punishment type, an undo path, a whitelist, per-role overrides, per-command limits and a dedicated log channel. That is the anti-nuke half of the problem, and none of it costs anything.

Anti-Raid is the paid half, at Basic — $4.99/month. It is two independent join-time defences on one configuration: raid detection counts how many accounts younger than a set age joined inside a rolling window and fires one action — mute, kick or ban — when the count crosses your threshold; Join Guard screens every join individually against up to 10 named conditions, each combining up to 5 rules with AND/OR logic, and runs whether or not a raid is in progress. Account age, avatar, banner and username shape are all available as rules, so the screen is expressible rather than a single toggle.

Two things sit around it that most security bots do not have. Dashboard Watcher records who changed which setting through the dashboard, free on Starter, 500 entries over 90 days across every system — and that is the exact gap named above, where Wick, Security Bot and Gaius all sell protection against a rogue administrator and none of them records that administrator editing the protection. Server Backup is a Max-tier feature ($7.99/month) covering channels, roles, server settings, bans, nicknames and role assignments, scheduled, restorable across servers and driven from the dashboard rather than from chat commands.

Everything else the same configuration surface reaches — 148 log event types, all 32 moderation commands, all 8 auto-mod rule types — is free, and it is one permission model rather than four. The full breakdown is in what Vetox’s moderation features actually do.

Which one to install

The most common good answer is two bots: Beemo for the flood, and something with anti-nuke for the insider. They do not overlap and neither replaces the other.

Frequently asked questions

What is the difference between anti-raid and anti-nuke? Anti-raid watches the door — it detects a burst of accounts joining and stops them. Anti-nuke watches people who are already inside and already trusted, and reverses mass deletions, mass bans and dangerous permission grants. A bot can be excellent at one and have nothing for the other, and most are.

Does Discord have built-in raid protection? No. Discord’s AutoMod filters message content with five trigger types and a cap of six custom keyword rules, and its verification levels gate everyone identically. Neither reacts to join rate, and neither can kick, ban or record a case. The whole third-party anti-raid market exists in that gap.

Is there a free Discord anti-nuke bot? Security Bot gives anti-nuke and anti-raid away and charges for chat automod instead. Wick’s own directory listing states its anti-nuke is free, and its premium card sells “Advanced” variants, which implies a free baseline. Vetox’s 20 protection rule types are free on Starter. Beyond those, anti-nuke is usually the paid half.

Can an administrator defeat an anti-nuke bot? Frequently, yes. Wick’s anti-nuke exempts anyone holding a role above Wick’s own, and Security Bot’s exempts anyone at or above its own role, so an insider who can move roles gets past both. Keep the bot’s role at the top of the hierarchy, restrict who can move roles, and run something that records configuration changes.

How many raid protection bots should I run? Two is a reasonable answer and three is usually one too many. Every extra bot is another permission model to audit and another set of logs nobody reads. Pick one general protection product and add Beemo, which is free and needs no configuration.

What should I check before buying one? Whether anti-raid is on the free tier or behind the paywall — it is the single most common surprise in this category. Then whether the bot records who changed its own settings, because an insider’s first move is to switch the protection off.

Method

Every capability, gate and price above was read from the vendor’s own site or documentation on 30 August 2026. Where two vendor surfaces disagree — Security Bot’s pricing page against its docs — both figures are shown and neither is presented as settled. Where a price is not published at all, it is recorded as unpublished. Absence claims are scoped to the sixteen moderation and security bots examined, not to Discord’s 10,121-app moderation category.