How to Stop a Discord Raid (2026 Playbook)

Pause your invites. Do that before you read the rest of this sentence — Server Settings, then Invites, then Pause Invites. It takes four seconds, it needs no bot, and it stops the raid from growing while you deal with the accounts already inside.

Almost every raid guide opens with a bot command. That is the wrong order. A raid is a flow problem before it is a cleanup problem, and the fastest valve is built into Discord and free.

I maintain a moderation bot, so I have a commercial reason to tell you otherwise. I am telling you this instead because it is what works, and because you will trust the rest of this guide more if I am straight with you about the first step.

In this guide

• Stop it right now: the first 60 seconds
• Use Discord's built-in tools first
• If you do not have the bot yet
• How to lock down a Discord server during a raid
• What actually counts as a raid
• Tuning detection so you do not kick real members
• Trap the next wave with honeypot channels
• Verification as a permanent gate
• Slow the chat instead of locking it
• When the raider is a staff account
• What this bot cannot do
• Clean up afterwards
• Your raid response checklist
• FAQ

Stop it right now: the first 60 seconds

Work in this order. Every step here uses only Discord's own features, so none of it depends on having a bot installed.

1. Pause invites. Server Settings, then Invites, then Pause Invites. Requires Manage Server. Existing links and your vanity URL stay intact, they just stop working, and this blocks Server Discovery joins too.
2. Raise your verification level. Server Settings, then Safety Setup, then Verification Level. Set it to High or Highest. High requires a verified email, an account older than five minutes, and ten minutes of membership in your server before anyone can talk.
3. Pause DMs. Open the server dropdown and choose Security Actions. This stops raiders from DMing your members, which is the part people forget — the worst damage in a raid is often the DM wave that follows it, not the channel spam.
4. Then start removing accounts and clearing channels.

That sequence matters. If you start banning first, new raiders keep arriving through the invite link and you are bailing out a boat without plugging the hole.

Only once the flow is stopped is a bot the faster tool, because banning forty accounts by hand is miserable and slow.

Use Discord's built-in tools first

Discord ships real raid defenses now, they are free, and several of them are better than what any third-party bot can do. Turn these on before you evaluate a bot.

A few details worth knowing, because the help pages bury them:

• Security Actions caps each pause at 24 hours. If you want a longer lockdown you have to re-apply it, so do not set it and assume you are covered for the week.
• Report Raid needs Ban Members or Kick Members, and it only exists on the desktop client in Community servers. Use it — Discord reviews these reports to build better tooling, and it costs you nothing.
• Admins and moderators bypass verification levels. Raising the level to Highest will not lock out your own staff.

Discord's own raid guide is genuinely good and worth ten minutes. So is the Pause Invites FAQ, the Security Actions article, and the breakdown of verification levels.

So where does a bot earn its place? Three jobs Discord does not do: punishing a join wave the instant it starts with a rule you chose, trapping raiders in decoy channels, and watching your own staff accounts for a permission grab. The rest of this guide is those three jobs.

If you do not have the bot yet

The commands below come from BetterAntispam, and slash commands only appear once a bot is in your server. If you are mid-raid and have none installed, do the native steps above first — do not stop to set up software while a raid is in progress.

When you are ready, it is about two minutes:

1. Click Add BetterAntispam to Discord
2. Pick your server, read the permission list, press Authorize
3. Type / in any channel to confirm the commands appeared
4. Run /setup and choose a log channel

You need Manage Server to add a bot, and Administrator to run most of the commands below. I wrote the full step-by-step version in the channel cleanup guide, including what to do when the commands do not show up.

Core anti-raid, anti-spam, and anti-nuke protection are all free. I will flag the paid parts as they come up.

How to lock down a Discord server during a raid

There are three different locks and people constantly reach for the wrong one. They do genuinely different things.

/lockdown silences channels

/lockdown channel denies Send Messages and Add Reactions to @everyone in that channel, and snapshots the previous permissions so unlock restores exactly what you had.

Two things it does not do, despite what you might assume. It does not hide the channel — people can still read it, they just cannot post. And it does not touch your server verification level.

/lockdown channel channel:#general reason:raid
/lockdown unlock channel:#general

Runner needs Administrator; the bot needs Manage Channels. Locking every channel at once and unlocking them all at once are Shield features. Locking a single channel is free, and in a raid that is usually all you need — raiders concentrate in whichever channel is busiest.

/serverlock closes the door

/serverlock action:enable

Read this next part carefully, because it is the single most misunderstood command in the bot. On its own, /serverlock action:enable sets a flag. It does not kick or ban anyone by itself. New members are only removed if you have separately turned on auto-kick or auto-ban, or if you set a password.

With a password, the bot creates a #server-password channel and new joiners have to type the password to get in. They get five wrong attempts before being kicked. Password mode is a Shield feature.

If you want the plain "nobody new gets in" behaviour, Discord's Pause Invites is honestly the better tool, and it is free. I would use /serverlock for the password gate specifically, not as a generic lock.

/raidmode is the blunt instrument

/raidmode

Raid mode punishes every new join, with no detection logic and no benefit of the doubt. It is what you want when a raid is actively in progress and you have accepted that a few real members will be caught.

Options cover the action, which punishment to apply, an enable_after delay, and whether to DM affected members.

One caveat that matters: raid mode lives in memory. If the bot restarts while raid mode is on, it comes back off. Do not leave it as your permanent protection.

[IMAGE: discord-lockdown-vs-serverlock.png — diagram showing which of the three locks blocks joins versus blocks posting. Alt: "Comparison diagram of Discord lockdown, serverlock, and raidmode showing lockdown blocks posting while serverlock blocks joining."]

What actually counts as a raid

By default the bot watches for a burst of joins in a rolling window of a few seconds, and it scales the threshold to your server size automatically.

That last part matters. Five joins in ten seconds is an obvious raid on a 40-member server and a totally normal Tuesday on a 40,000-member one. A single fixed number is wrong for almost everybody, so auto-scaling is on by default.

Two thresholds, not one

The "considers" and "certain" columns behave differently, and this is the design decision I am most willing to defend.

Hitting the certain threshold punishes the wave immediately, no questions asked.

Hitting the lower considers threshold only triggers a response if corroborating signals add up. The bot scores what it sees, and needs a combined score of 3 before acting. Roughly:

• Join rate well above your normal — worth 3 on its own
• Accounts created in the last 24 hours — worth 3
• Repeated or near-identical usernames — worth 2
• Everyone arriving on the same invite — worth 2

So six joins in ten seconds on a 900-member server does nothing if they look like six ordinary people. The same six, all registered this morning and all on one invite link, gets stopped.

What it does when it fires

The default punishment is quarantine — roles stripped, quarantine role applied — which I think is the right default because it is the only reversible one. If the role is missing, it falls back to a kick.

You can switch it to kick, ban, mute, or captcha. Two specifics: mute is a fixed 10-minute Discord timeout, and captcha drops the joiner into an #antiraid-captcha channel with a 120-second default to solve it (adjustable from 10 to 600 seconds). Fail it and they are kicked.

Raid mode then clears itself automatically after 10 minutes with no further joins, so you are not left locked down after the wave passes.

You can exempt specific users and roles from all of this with the whitelist, which is worth setting up for your staff before you need it.

Tuning detection so you do not kick real members

Set the sensitivity to match how bad a false positive is for you, then leave the thresholds alone.

Three presets multiply every number in that table:

My honest advice: leave it on balanced unless you have evidence. I see far more servers broken by over-tightened thresholds than saved by them. A strict setting on a server that gets a legitimate traffic spike — a streamer shoutout, a Reddit post landing — will quarantine a wave of genuine new members, and they will not come back.

If you do get a false positive, the fix is usually the whitelist or a longer window, not a lower threshold.

There are also individual join checks you can enable: a minimum account age, kicking accounts with no avatar, catching usernames that contain invite links, and a "suspicious account" score with presets. These are off by default. Turn them on one at a time so you can tell which one caused a complaint. Fine-grained control over the similar-name detector and individual suspicion factors are Shield features.

Trap the next wave with honeypot channels

A honeypot is a decoy channel that no real member has a reason to post in, so anything that posts there has outed itself.

This is the highest-signal anti-raid tool I know of, and almost nobody uses it. Detection based on join patterns is guesswork. A honeypot has essentially no false positives, because a human reading your channel list does not wander into a staff-looking channel and start typing.

Set it up with /honeypot settings (runner needs Manage Guild). It is off by default. The bot can auto-create bait channels for you, and it deliberately uses staff-style names rather than obvious "free-nitro" bait — self-bots increasingly skip the obvious traps.

The default action is a softban: a ban that removes the account's messages from the last seven days, followed immediately by an unban. The account is gone along with its spam, but it is not permanently blocked, which is the right call for what might be a compromised friend's account.

By default staff are ignored, so an admin poking around does not softban themselves. You can also have it purge that account's messages server-wide from a recent window.

Voice-channel traps are a Shield feature; text traps are free. Chat inside a decoy voice channel counts too.

Verification as a permanent gate

Verification stops raiders before they can post anything, and it is off by default — so this is an opt-in you have to actually go turn on.

Run /verification settings. New members get an unverified role and have to clear a challenge: a button click, a captcha, or a simple maths question. They get 3 attempts, and failing or timing out kicks them by default (you can switch that to a timeout, a quarantine, or nothing at all).

The option worth enabling is verification lockdown, which denies the unverified role View Channel, Send Messages, Connect, and Speak across the server. Your verification channel and any honeypot traps stay reachable on purpose. Without it, unverified accounts can still read everything.

Be honest with yourself about the tradeoff: every gate costs you real members. Some people will not bother. Discord's own Rules Screening is lighter-touch and free, and for a small friendly server it is the better choice. Reach for a full verification gate when you are being targeted repeatedly, not by default. See how it fits with the rest of the stack on the verification section of our homepage.

Slow the chat instead of locking it

Slowmode keeps a channel usable while making spam pointless, which a lockdown does not.

/slowmode duration:10s
/slowmode duration:off

Accepts anything from 0 up to 21,600 seconds (six hours), takes Manage Channels, and applies to the channel you run it in. It is free.

This is my preferred move for a moderate spam wave. Thirty seconds of slowmode makes flooding useless, and your actual community can still talk — which matters, because a locked-down server during an incident is where people go to panic in the absence of information.

/slowmodebulk applies slowmode to up to five channels from one wizard. That one is a Shield feature. Worth knowing so you are not hunting for it: the single-channel version is free forever.

When the raider is a staff account

Anti-nuke watches for destructive administrative actions and stops the account doing them, which is a completely different threat from a join raid.

A join raid is loud and annoying. A compromised moderator account deleting every channel is existential. Run /antinuke settings — it is off by default, which I would call the most important thing in this article that you probably have not turned on.

It counts actions per person in a rolling 10-minute window. Some of the defaults:

The single-action thresholds are the interesting ones. One unverified bot added, one Administrator grant, one vanity URL change — because there is no innocent reason for those to happen in bulk, and adding a malicious bot is how most server takeovers actually start.

The default response for a person is quarantine. For a bot it strips permissions and kicks. For mass channel creates and deletes it reads the audit log and undoes them.

Unlike the anti-raid side, anti-nuke reports its own failures into your log channel, so you find out if it could not act. More on how this layer works on the security section of our homepage.

What this bot cannot do

Every guide like this ends with a pitch. Here is the honest list first, because you should know what you are and are not covered against.

• A slow raid walks straight past it. Detection is burst-based, over windows of 5 to 15 seconds. Accounts trickling in one every thirty seconds will never trip a threshold, no matter how many arrive. Verification and honeypots are your defence there, not join detection.
• Anti-raid punishment failures are silent. If a ban or kick fails because of role hierarchy, the anti-raid side does not tell you. Anti-nuke does. Check that the bot's role sits above the roles it needs to action.
• Raid mode and join counters are in memory. A bot restart resets them.
• Automatic detection does not lock your channels. When a join wave triggers, the bot punishes the joiners. It does not run /lockdown for you. That is a deliberate call — automatically silencing a whole server on a threshold would be worse than the raid — but it means a human still has to make that decision.
• There are no anti-raid transcripts. You get log embeds and moderation cases, not a saved transcript of the raid. Transcripts exist for spam mutes and for channel purges, not join waves.
• Anti-nuke depends on the audit log, which is rate-limited and occasionally slow. A very fast actor can get more done than the thresholds imply.
• Discord's limits are still Discord's limits. Timeouts cap at 28 days, bulk message deletion cannot touch anything older than 14 days, and every action is subject to Discord's rate limits.

None of that makes the tool useless. It does mean "install a bot" is not a strategy on its own, and anyone who tells you otherwise is selling.

Clean up afterwards

Once the flow is stopped and the accounts are gone, you still have channels full of spam.

Do not clone the channel to clear it, which is the advice you will find everywhere — it silently breaks every webhook and bot setting pointed at it. Use a purge instead, and keep the transcript. The channel cleanup guide covers the whole thing, including why 14-day-old messages delete so slowly.

Two things specific to post-raid cleanup:

• Keep the transcript before you wipe. The spam is worthless but the pattern is not — which accounts, in what order, on which invite. That is what tells you how they got in, and it is what you attach to a report. A purge log gives you this; deleting the channel does not.
• If you deleted something you needed, /snipe can often bring it back. There is a section on recovery in that guide.

For channels that never need history — bot commands, notification dumps, the bait channels themselves — put them on a schedule with auto-nuke and stop thinking about them.

Then do the part everybody skips: work out which invite link they came in on, and revoke it.

Your raid response checklist

That last one catches people out. Servers stay quietly locked down for weeks after an incident and nobody notices growth has stopped.

[IMAGE: discord-antiraid-dashboard.png — the /antiraid settings dashboard showing sensitivity and punishment options. Alt: "BetterAntispam antiraid settings dashboard with balanced sensitivity and quarantine punishment selected."]

Get protected before the next one

If you are reading this after a raid rather than during one, the useful window is right now, while you still remember how it felt.

1. Set your verification level and turn on Discord's Raid Protection and AutoMod — free, native, do this first
2. Add BetterAntispam and run /setup with a log channel
3. Turn on /antinuke settings, because it is off by default and it covers the worst case
4. Add one honeypot channel with /honeypot settings
5. Whitelist your staff roles so nothing catches them mid-incident

Anti-raid detection, anti-nuke, text honeypots, verification, single-channel lockdown, and /slowmode are all free. Server-wide lockdown, voice traps, password gating, and /slowmodebulk are the Shield extras. You can see the raid-specific features on the anti-raid section of our homepage.

And set up that log channel before you need it. Every raid post-mortem I have seen that went nowhere went nowhere because there was no record.

FAQ

What is a Discord raid?

A coordinated flood of accounts joining a server to disrupt it — spamming channels, DMing members, or getting the server reported. It is distinct from a nuke, where someone with permissions destroys the server from the inside.

How do I stop a raid immediately?

Pause invites in Server Settings, raise your verification level to High, and pause DMs through Security Actions. All three are native, free, and take under a minute combined. Remove accounts and clear channels after the flow is stopped.

Can I stop a raid without a bot?

Yes, and for the first minute you should. Pause Invites, verification levels, Security Actions, AutoMod, and Discord's own Raid Protection cover the immediate response. A bot is for automating the removals and for threats Discord does not watch, like a staff account being used to nuke the server.

How many joins count as a raid?

It depends on your size. Roughly 3 joins in 10 seconds on a 50-member server, about 15 on a 10,000-member one, and 100 or more on a very large server. The thresholds scale automatically and a lower "considers" tier only fires when other signals back it up.

Will anti-raid protection kick my real members?

It can, which is why the default sensitivity is balanced rather than strict and the default punishment is a reversible quarantine. Whitelist your staff roles, and do not tighten thresholds without evidence — a legitimate traffic spike looks a lot like a raid.

What is the difference between anti-raid and anti-nuke?

Anti-raid watches people coming in. Anti-nuke watches people already inside with permissions — mass channel deletions, Administrator grants, adding unverified bots. They solve different problems and anti-nuke is off by default.

Does locking down my server stop new members joining?

No, and this trips people up constantly. /lockdown stops people posting; they can still join and still read. To stop joins, pause invites or use /serverlock.

Is Discord's built-in Raid Protection enough?

For many servers, honestly, yes — combined with AutoMod and a sensible verification level. Turn those on first. Add a bot when you need punishments on your own rules, honeypot detection, or protection against an insider with permissions.

How do I find out how the raiders got in?

Check which invite link the joins came through, then revoke it. Your join logs record this, which is the argument for setting up a log channel before an incident rather than after.