Not just a filter. A whole server crew.
Johnny catches slurs no word list can read, keeps quiet channels talking, welcomes people and hands out roles, runs tickets and ban appeals, and locks the door during a raid. One dashboard for all of it, and a filter you can test before you install.
Free to add.
Text inside shrug.png matched a blocked term.
Read by OCR, 98% confidence. Author warned.
Blocked term hidden with leetspeak and zero-width spaces.
Matched ▮▮▮▮▮▮ after normalising. Second strike, muted 10 min.
@mod marked the 14:02:34 match as a false positive.
Recorded. The filter adjusts for this server.
What it does
Most servers run four bots to get this. Johnny does it from one place, with one log and one dashboard.
Leetspeak, spaced letters, lookalikes, reversed text, zero-width tricks, and text inside images. Reads screenshots, so a fake crypto giveaway posted as a picture gets deleted too.
Chatter learns how a channel talks and posts a message of its own when things go quiet. Ping someone's role and a picture appears. Popular messages land on a starboard.
Welcome and goodbye messages, roles on join, roles from a reaction. Set up on the dashboard in a minute.
Tickets with saved transcripts, a ban appeal link the bot sends automatically, and a safe word anyone can type to reach staff quietly.
Mass bans, channel wipes and role purges get stopped and reversed. Raids trigger slowmode. Brand-new accounts get flagged. Invite, phishing and scam links go.
Auto-replies to common questions, weekly reminders, a card maker for announcements, a FAQ command, and a dashboard that shows what your server actually uses.
The filter
Most bots stop at a word list. Johnny checks the places people actually hide things.
Turn OCR on and every attachment, sticker and embed thumbnail is read, so a slur on a screenshot is treated like a slur in chat.
Leetspeak, spaced-out letters, lookalike characters, zero-width joins and stretched vowels are all normalised before matching.
Messages are translated before they are checked, so your word list works whether someone writes in English, Spanish or Tagalog.
When enabled, an AI pass reads the whole sentence before anything is deleted. Quoting a rule, reporting an insult or discussing a word does not get someone punished.
Mass bans and role deletions are reverted in seconds, wiped channels are recreated, and the account that did it is banned.
Dismiss a false positive from the log and the learning engine adjusts for your server, so the same match gets less likely each time.
One rule
Add a word once. Johnny matches every way people try to sneak it past a filter.
All six resolve to the same rule. You maintain one word list, not fifty variants of it.
Anti-nuke · Premium
Mass bans and role deletions are reverted in seconds, wiped channels are recreated, and the account that did it is banned.
Detects a compromised account or malicious admin trying to wreck the server. Each is counted in a sliding time window; crossing a threshold triggers an alert.
Setup
One click. It asks for 13 permissions, all listed in the FAQ below, and nothing it does not use.
Run /setup to pick a preset, or choose your words, a sensitivity and a log channel yourself. Dashboard or slash commands, your call.
Everything it does shows up in the log. Dismiss a false positive with one click, and it adjusts.
Configure it without leaving Discord. Each command is permission-gated, so only the roles you choose can change rules or act on members.
Warnings are kept per member with a full history, and the filter escalates repeat offenders on its own.
Pricing
Every moderation feature is free on every server. Premium adds the parts that cost money to run.
Run /premium in your server first. It shows the claim note to paste into Ko-fi so the payment lands on the right server.
FAQ
Whitelist a phrase and it never trips the filter again. If something does get caught, one click on the log entry dismisses it and the filter learns. With Premium, an AI context check reads the whole sentence before anything is deleted, so quotes and reports are allowed through automatically.
Click Dismiss as FP on the log entry. The dismissal is recorded and the learning engine adjusts for your server, so the same false positive gets less likely each time. Discord does not let anyone restore a deleted message, so the member is told why it was removed.
Each action is logged with who, which rule, when, and a copy of the flagged message so your mods can review it. Violation records are deleted after 90 days. Messages that pass the filter are never stored, except inside a ticket, where the transcript is saved when the ticket closes. Full details are in the privacy policy.
View channels, send messages, read message history, manage messages, embed links, attach files, moderate members, kick members, ban members, manage roles, manage channels, manage webhooks and view audit log. The last three are what anti-nuke uses to identify the attacker, recreate wiped channels and remove rogue webhooks.
The filter, anti-spam, warnings, lockdown, tickets, appeals, welcome messages, roles and every moderation command are free on any server. Reading text inside images, translation, the AI context check, ping images and anti-nuke are Premium, $8 a month per server. Before paying, run /premium in your server and paste the shown Server ID ... claim ... note into Ko-fi. Payments without that note cannot be matched to a server.
Two minutes to set up. Free for every server.