Protection
Defend the server against raids and rogue staff accounts
"Protection" is really seven independent shields bundled under one module, plus server backups. You don't have to run all of them — most servers start with anti-nuke and backups, then add the rest as they need it.
Turning it on or off
Every server chooses for itself — switch Protection on or off from your dashboard, under Moderation & Safety. When it's off: raid and mass-deletion attacks will not be blocked automatically.
Setting it up
- 1
Build your trusted list first
Add every genuinely trusted staff member, role and bot to Anti-Nuke's trusted list before turning it on — otherwise your own team's routine cleanup work (bulk channel changes, mass role edits) can trip the same alarms meant for an attacker.
- 2
Turn on Anti-Nuke with sane limits
The defaults are a reasonable starting point; tighten the windows only once you've confirmed normal staff activity doesn't false-positive.
- 3
Take a backup while things are calm
A backup taken mid-raid is a backup of the mess. Schedule one and also grab a manual one whenever you make a big structural change.
- 4
Decide on Join Gate and Bot Gate separately
They run independently — you can screen humans without touching how new bots are handled, or the reverse.
- 5
Set a Raid Mode auto-trigger if you're a public, high-traffic server
Smaller invite-only communities usually don't need it and can rely on the manual trigger.
Server backups
A backup is a full snapshot: channels, categories, permission overwrites, roles, and forum tags. /backup create takes one on demand; scheduled backups run daily, weekly or monthly automatically.
Restoring gives you two modes: safe (additive — recreates what's missing without touching what's already there) or full (wipes the server structure and rebuilds it exactly as the backup recorded it). Full restore is the one to be careful with.
Retention limits — keep the last N backups, or backups newer than N days — only ever prune backups your schedule created automatically. A manual backup you triggered yourself is never auto-deleted.
What each shield does
Anti-Nuke
Watches for the actions a compromised staff account or a malicious bot would take fast: mass channel or role deletion/creation, mass kicks or bans, stripping roles from members, creating webhooks, escalating permissions, adding a bot, or changing server settings. Each of those has its own rate limit (a max count within a time window) and its own response — remove the offender's roles, kick, ban, quarantine them, or just notify you. Trusted users, roles and bots bypass every check.
Join Gate
Screens human members as they join: minimum account age, blocking accounts still on Discord's default avatar, and a username pattern you define. Each check can respond differently — do nothing but log it, kick, ban, or quarantine into a holding role until a moderator reviews them.
Bot Gate
A separate, bot-specific version of the join gate (the human checks above explicitly skip bots). Run it as an allowlist — only bots you've named can join — or in block-all-new-bots mode, where every new bot addition gets flagged for review.
Raid Mode
Can trigger itself automatically once joins per window cross a threshold you set, or you can flip it on by hand. While active it tightens auto-moderation across the board (a strictness multiplier), can lock the server down immediately, and posts to an announcement channel. It disarms itself automatically after a time you choose, or you disarm it manually.
Panic Mode
The one shield that's always manual — there's no auto-trigger, because it's meant for "something is badly wrong right now." It locks the server the same way Raid Mode does, but also revokes every active invite link, so a raid already spreading through a shared invite can't keep pulling people in while you deal with it.
Alt-Account Heuristics
Scores each new join using account age, default avatar, whether the account has a prior ban or tempban case in your server's own history, and whether Invite Tracker flagged the join as coming from a fake/suspicious invite. Once the score crosses a threshold, it takes the action you've set — including quarantine, so a human still gets the final say.
Webhook Audit
Alerts when a webhook is created that isn't on your trusted allowlist — a common way nukers post malicious content without needing bot permissions. Optionally also alerts on webhook edits and deletions.
Commands
Tips & things people get wrong
- An empty trusted list is the single most common cause of "Anti-Nuke banned my own moderator" reports — fill it in before you enable anything.
- Panic Mode's invite revocation is one-way in the sense that old invite links won't work again — recreate the ones you actually want to keep after you disarm it.
- Alt-Account Heuristics gets meaningfully better once Invite Tracker has been running for a while — the "fake invite" signal has nothing to work with on a brand new server.