The raid hit at 3am. Nobody was awake.
Two hundred accounts join in ninety seconds and start walking through your channels. Wardn is watching the join curve, not waiting for a mod to notice the ping.
<5s
from first signal to response
24/7
including your worst timezone
1 tap
to reverse the whole event
By the time you react, it's already screenshotted.
It starts while your team sleeps
Raids are scheduled for when a server is thinnest. That's exactly when your human coverage is zero.
Mass-join looks normal for about 40 seconds
Then it doesn't. The window between 'busy day' and 'we're being raided' is short, and nobody's staring at the join feed.
Someone gets a webhook and nukes channels
One compromised role, one careless invite, and your history is gone in a minute.
Manual lockdown takes too long
Find the setting, pick the roles, apply per channel — meanwhile the raid is still typing.
Cleanup afterwards is worse than the raid
Hundreds of messages, hundreds of accounts, and the risk of banning a real member caught in the sweep.
It happens again next weekend
Same crew, new accounts. Nothing you did last time carried forward.
Ninety seconds of chaos, contained in four.
And because every action is reversible in one tap, a false alarm costs you nothing but a lockdown notice.
Every action comes with a reason.
Not a confidence score. A sentence your mod team can read, argue with, and overturn in one tap.
Tuned to your server, not the average server.
A meme that gets you banned in a corporate Discord is just Tuesday in a gaming one. Wardn learns which one you are — from your team's own calls.
It would rather ask than over-ban.
Borderline cases go to your mods with context attached, instead of quietly removing a real member.
It reads everything a mod would look at.
Most bots only see the words in a message. The stuff that actually gets past them is in a picture, behind a link, or sitting in someone's display name.
Tone, intent and context — including the sarcasm, the in-jokes and the borderline calls your server actually allows.
Screenshots, memes, QR codes and text baked into pictures. Wardn looks at the image itself, not the filename.
Where a link really goes: look-alike domains, redirect chains, fake giveaway pages and drainer sites.
Usernames, nicknames and display names — offensive handles and names used as free advertising, caught on join.
The stuff that gets through is never plain text.
Anyone who spams for a living knows keyword filters read words. So they post a screenshot, a QR code, or a link that looks fine until you follow it. Wardn opens the picture.
Caught on the picture itself, hidden instantly, and logged so a mod can confirm the call without the whole channel seeing it first.
Twitch, YouTube and "DM me for cheap" ads posted as a screenshot to dodge link filters. Wardn reads the words inside the picture.
The code gets decoded and the destination checked — the same way one of your mods would if they had the time and the patience.
Redirect chains, fake giveaway pages and one-character-off domains, resolved before your members click them.
Same raid crew. Very different night.
Right now- ✕200 accounts through every channel before anyone wakes up
- ✕Manual channel-by-channel lockdown while it's ongoing
- ✕Hours of cleanup and a bloated ban list the next day
- ✕Real members swept up in the mass-ban
- ✕The same crew back next weekend
Two weeks in- ✓Join-curve anomalies flagged in seconds, day or night
- ✓One-command lockdown that leaves regulars untouched
- ✓Cleanup already done, with a reviewable log
- ✓Verified members explicitly protected from sweeps
- ✓The pattern remembered, so round two lands even flatter
It has to earn write access. Slowly.
Every horror story about AI moderation starts the same way: someone gave it permissions on day one. We don't. Here's the whole process.
- 01nothing installed
We talk to your mod team first
Before anything gets installed, we sit down with the people who actually do the banning. What sets you off, what you let slide, where the line is on this server.
- 02no write access
Read-only for a week
Wardn joins with zero powers and just watches. No deletes, no warns, no timeouts. It's gathering what your server normally sounds like.
- 03your rules, not ours
You show it your calls
Your mods point at what they banned, warned, or ignored — and why. That's the training signal. Not a generic filter list. Your standards.
- 04still nothing live
Sandbox run
We replay your real history through the trained system. Every flag, the action it would've taken, the reason, the member, and their prior warnings — all laid out.
- 05you approve
Your team reviews it
You read the whole sandbox before a single permission is granted. Too aggressive on memes? Tune it. Missed the crypto DMs? Teach it. Zero risk, because nothing has happened yet.
- 06your call
Then, and only then, it goes live
Activation is your call, one channel or all of them, with the scope you choose. And every action stays undoable.
You never have to take our word for it.
By the time Wardn can delete a message, you've already read every decision it would have made on your own server.
See everything. Undo anything.
Moderation you can audit at a glance — plus the same commands right inside Discord, because nobody wants to open a browser mid-raid.
- 2m agoScam link@nitro_gifts_xBanned
- 8m agoSelf-promo@beatsbyleoRemoved + DM
- 14m agoSpam@lex_09Slowmode nudge
- 21m agoToxicity@mikathecatWarned
- 37m agoScam link@steamtrade_botBanned
- 52m agoClean@rinNo action
Mod team audit log
- @kira banned @tradebot_44 — 11:02
- @sol reversed a Wardn timeout — 09:48
- @kira approved rule “no DM trading” — Tue
- @dax muted @grifter — Mon
Mods are accountable too. Everything is logged, both directions.
In Discord, too
/wardn history @lex_09
/wardn undo last
/wardn lockdown 10m
Most moderation bots were built for a rulebook. Yours isn't one.
The old bots match strings. The new ones are aimed at developers and token communities. Neither of them is built for a mod team of six volunteers running a 40,000-member gaming server.
Word lists, regex, thresholds. Great until the spam changes one character — then it's your weekend.
Powerful, configurable, and written for people who enjoy YAML. Your mod team does not enjoy YAML.
Trained per server on your own moderation history, reviewed in a sandbox before it gets a single permission, and explained in plain language.
What mod teams actually ask us.
What if it bans someone it shouldn't?
Every action is reversible in one tap, from the dashboard or a slash command. And the reversal is a training signal — it won't make that call again.
Do we lose control of our own server?
No. You set the scope, you approve every new pattern, and your mods can override anything. Wardn does the volume; you keep the judgment.
How long until it's actually useful?
Usually a week of watching, then a sandbox review, then activation. Most teams see the difference in the first few nights.
What does it cost?
Every server gets its own quote. Message volume, how you want it set up and how deep the integration goes all change the work involved — so there's a one-off setup fee plus an ongoing fee that covers running it. You'll get exact numbers on the first call, before anything is installed.
Why can't we start next week?
We onboard in batches of five servers at a time, because each setup is trained and tuned by hand. When a batch is full, the next one waits. Reaching out early is how you get into the current batch.
We onboard five Discord servers at a time. That's it.
Wardn isn't a bot you invite and forget. Each server is trained on its own moderation history and tuned by hand with your team, so we take on five at a time and finish them properly before the next batch starts.
When the current batch is full, the next one waits. If your nights are bad now, early is the only way in.
No public price list — and that's deliberate.
A 4,000-member server with a quiet chat and a 90,000-member server pushing a hundred thousand messages a day are not the same job. Your offer is built for your server.
What shapes your quote
- ◆Message volume — how much traffic we're reading every day
- ◆How you want it set up: channels, roles, escalation, tone
- ◆Depth of integration with your existing mod stack and logs
- ◆How much moderation history there is to train on
A setup fee, then an ongoing fee.
The setup fee covers the training, tuning and sandbox review for your server. The ongoing fee covers actually running it — the models, the watching, the support. Both are quoted up front, in plain numbers, before you commit to anything.
Been raided? Tell us how it went.
The more specific you are about what got through, the more useful our first conversation is.
Keep reading
Discord moderation bot →
What a moderation bot has to get right in 2026 — and where the classics stop.
AI Discord moderation bot →
The training loop: every override your mods make becomes a rule you never write.
Discord AutoMod alternative →
Why wordlists lose to one swapped character, and what reads intent instead.
Discord scam & phishing protection →
Unmasking redirect chains, fake giveaways and QR drainers before members click.
Discord self-promotion bot →
Telling the artist sharing work apart from the person farming your members.
Discord content moderation →
NSFW, gore and slurs handled the same way at 3am as at 3pm.
Discord raid protection →
Lockdown that doesn't punish every real member who happened to join today.