Docs / Core Services / Sentinel
Built in Service id sentinel 30 commands 32 settings

Sentinel

A rule engine with twenty-one conditions and thirteen actions, anti-nuke that reverts the damage, raid detection that spots look-alike accounts and the invite they came through, lockdowns with timers, server change logs, Discord's own AutoMod managed from the same dashboard, and a case log with appeals. Every action leaves a numbered case with the evidence that caused it.

Turn it onDashboard, Services, Sentinel, switch to Enabled
First stepRun setup
PlanoEvery feature included for every server: unlimited automod rules, unlimited conditions per rule

Set it up

The dashboard's health check expects: Quarantine role, Moderator roles. It shows a warning while one of these is missing or points at something deleted. The steps below say what each one does.

  1. Turn the service on. On the dashboard, open your server's Services page and switch on the Moderation card. Until it is on, every moderation command answers "Sentinel is not enabled for this server." and the automod does nothing.

  2. Open Sentinel and pick Setup, the chip under Ops in the service's navigation. Only dashboard managers of the server see it; the page tells anyone else that Manage Server permission is needed. The "What Auto-Setup Does" list on that page is the plan for the run.

  3. Wait for the Readiness check at the top of the page. It confirms "Bot is in this server", "Can create the log channel (Manage Channels)" and "Can create the Quarantined role (Manage Roles)", then says "Ready to run setup." Grant the bot any permission marked with a cross before you continue, or the run stops at that step.

  4. Choose a protection level: Relaxed (spam control and known phishing links), Standard (preselected, balanced protection) or Strict. Leave "Enforce rules immediately" unchecked. The rules then start in log-only mode: the bot records what each rule would have done without touching anyone, so you can watch them before they enforce.

  5. Click Run Auto-Setup and confirm. In order, the wizard: checks what already exists; reuses a category with "mod" in its name or creates a private "Moderation" category; creates #sentinel-logs and makes it the moderation log channel; creates a "Quarantined" role with no permissions and denies it sending messages, reacting and using threads in every text, announcement and forum channel; detects moderator roles (roles named with mod, admin, staff or moderator that hold Kick Members, Ban Members or Manage Messages); saves those settings; creates #message-log and #server-log, points the message log and the server logs at them and turns them on; turns on anti-nuke alerts, escalating timeouts and the message log; installs the rules of your level; and posts a "Sentinel Moderation Active" card in #sentinel-logs.

  6. Read the Setup Result page. Every step shows done, skipped or an error with its reason. An error is almost always a missing bot permission; fix it and run setup again. Steps that succeeded are not repeated, and channels or roles that already exist are reused.

  7. Open Settings and check Roles & channels. If the wizard found no moderator roles, add them here: until you do, only Administrators can run the staff commands. Confirm the moderation log channel and the quarantine role. Under Moderator commands, set a default timeout if you want /timeout to work with the duration left empty, and decide whether you want anonymous moderation and the Confirm prompt before destructive actions.

  8. Try it. Run /warn on a test account with a reason. You should see "Warning issued to @member. Case #1", the test account should receive the warn DM, and a "Warning Issued" card should appear in #sentinel-logs. Run /history on the same account to see the case and the warning count.

  9. Confirm on the dashboard. The Cases page lists case #1 and the Overview page shows the protection strip with the raid and lockdown controls. After the log-only rules have run for a while, open each rule on the Rules page and untick "Dry Run (log only)" to start enforcing it. The "Dry-run all rules" switch under Settings > Warning points & retention overrides every rule at once, so leave it off.

If the Setup page also shows "Or describe it", you can describe your community and how strict you want moderation, preview the rules drafted for you, and apply them instead of a preset.

1 optional features with their own switch
  • Moderation log channel: moderation log

Como funciona

  1. Run setup. Pick Relaxed, Standard or Strict. The wizard creates the mod log, the message log and the server log as private channels, a Quarantined role, finds your moderator role, installs rule templates in dry-run mode and turns on the recommended protection.
  2. Rules watch the server. Each rule combines conditions, an action list and scoping: which channels, which roles are exempt, and whether it fires on a message, a join or a rename.
  3. Every action becomes a case. Warnings, timeouts, kicks, bans and quarantines, automatic or by command, get a numbered case with a why-this-fired panel.
  4. Members appeal, staff decide. /appeal or the web appeal page, reviewed in the Appeals tab. Overturning a case reverses it.

Permissions the bot needs

Generated

The invite link grants these. If one is missing the service page in the dashboard shows a re-grant banner, and the affected commands say what they cannot do instead of failing silently.

Kick MembersBan MembersManage ServerView ChannelSend MessagesManage MessagesEmbed LinksManage NicknamesManage RolesTimeout Members

What is included

GeneratedSee pricing

Every Sentinel feature is included for every server, on the shared bot and on custom bots. A subscription adds a custom bot identity and marketplace access, not features here.

RecursoIncluded
OverviewSim
CasesSim
Case Management (notes, status)Sim
User HistorySim
View RulesSim
Create/Edit RulesSim
View AppealsSim
Review AppealsSim
View ReportsSim
Review ReportsSim
Moderation AnalyticsSim
Mod Team InsightsSim
Rule TemplatesSim
Auto-Setup WizardSim
View SettingsSim
Edit SettingsSim
Limites
Automod RulesUnlimited
Conditions per RuleUnlimited

Comandos

GeneratedEvery command in detail
comandosO que isso fazWho by default
/appeal
/appealAppeal a moderation case against youEveryone
/ban
/banBan a member from the serverModerator roles
/case
/caseLook up a case by its numberEveryone
/history
/historyView moderation history for a memberEveryone
/kick
/kickKick a member from the serverModerator roles
/lockdown
/lockdownLock this channel, the whole server, a role, new joins or invitesModerator roles
/note
/noteAdd an internal staff note to a member's recordModerator roles
/purge
/purgeBulk delete messages in this channelModerator roles
/quarantine
/quarantineQuarantine a memberModerator roles
/raidmode
/raidmodeSet raid mode level (owner only)Moderator roles
/report
/reportReport a member to the moderation teamEveryone
/slowmode
/slowmodeSet slowmode in this channel (0 to disable)Moderator roles
/softban
/softbanBan then immediately unban: clears recent messages, the member may rejoinModerator roles
/temp-role
/temp-roleGive a member a role for a limited timeModerator roles
/timeout
/timeoutTimeout a memberModerator roles
/unban
/unbanUnban a user by IDModerator roles
/unlockdown
/unlockdownLift a lockdown: this channel, the server, a role, joins or invitesModerator roles
/unquarantine
/unquarantineRelease a member from quarantineModerator roles
/untimeout
/untimeoutRemove a member's timeoutModerator roles
/voicemod
/voicemod muteServer mute a member in voiceModerator roles
/voicemod unmuteLift a member's server muteModerator roles
/voicemod deafenServer deafen a member in voiceModerator roles
/voicemod undeafenLift a member's server deafenModerator roles
/voicemod disconnectDisconnect a member from voiceModerator roles
/voicemod moveMove a member to another voice channelModerator roles
/warn
/warnIssue a warning to a memberModerator roles
/watchlist
/watchlist addAdd a member to the watchlistModerator roles
/watchlist removeRemove a member from the watchlistModerator roles
/watchlist listShow who is on the watchlistModerator roles
Report Message
Report MessageRight-click a message and pick Apps, then Report Message.Everyone

Everything it does

Conditions: content

Ten conditions about what a message contains. Defaults shown; every number is adjustable.

Word and phrase filter. Contains, exact or regex, with an anti-evasion normaliser that folds look-alike characters, leetspeak and spaced-out letters.
Message spam rate. Too many messages too fast. Default 5 in 10 seconds.
Mention flood. Too many mentions in one message. Default 5.
Caps. Percentage of a message in capitals, default 70, checked once it has 8 letters.
Attachment flood. Too many files in a window. Default 4 in 30 seconds.
File type block. By extension, such as .exe.
Duplicate messages. The same message repeated. Default 3 in 60 seconds.
Emoji spam. Default 10 in one message.
Newline spam. Walls of text. Default 8 line breaks.
Sticker spam. Default 3 in one message.

Conditions: accounts, names and links

Eleven conditions about who is posting and where a link leads.

Links. Block all links, only Discord invites, or a domain list you write, with an allow and block list for invite codes.
Freshly registered domains. Flags links to domains registered in the last 30 days, looked up live and cached for a week.
Known phishing links. Matches a shared list of known phishing domains pulled from two public feeds and refreshed every 6 hours, plus the domains you block or allow for your server.
Links in profile names. Catches advertiser accounts with an invite or URL in their name.
Name filter. The word filter applied to the display name, the username or both, on join and on rename, with contains, exact, regex and a wildcard mode (* and ?).
No avatar. Matches members without an avatar on join or rename, so a rule can act on it outright. A No avatar on join template installs switched off.
Staff impersonation. A name one edit away from a staff member's, or the same avatar.
New accounts. Accounts younger than a number of days. Default 7.
Author roles. Limit a rule to members who hold, or lack, chosen roles.
Risk score and risk band. Fire only for members whose risk score or band is high enough.

Actions and how rules run

Thirteen actions a rule can take, in the order you list them.

Warn. Logs a case and DMs the member with a hint on how to appeal.
Timeout. From 1 second to 28 days, 10 minutes by default. Repeat offenders can get longer timeouts automatically.
Kick and ban. Bans can delete up to 7 days of messages, expire on their own as temp bans, and copy to your other servers.
Quarantine. Applies your quarantine role. Release by command or from the member's page.
Voice mute, deafen, disconnect or move. Timed mutes and deafens lift themselves, up to 28 days. Skipped quietly if the member is not in voice.
Delete, alert, DM, rename. Delete the message, post an alert to a channel, send a custom DM with {server}, {rule}, {reason} and {user}, or strip banned words and zalgo from a nickname.
Scoped and ordered. Rules run by priority, only in the channels you allow, never for exempt roles, and match on any condition or all of them.
Escalation and a cooldown per rule. Any action can wait until a member has a number of warnings. Each rule has its own punishment cooldown per member, 10 seconds by default, 0 to 3,600, so one burst cannot stack.
Dry run. A single rule, or the whole server, can log what it would do without doing it.
Fifteen templates. Anti-spam, anti-raid, invite control, mass mentions, new accounts, no avatar on join, phishing, hate speech, fresh domains, staff impersonation, advertising accounts, nickname sanity, executable attachments, stickers and walls of text, and high-risk containment. Save your own too.

Anti-nuke

For the night a staff account is hijacked. Twelve watches, each with its own count and window, and a response that undoes the damage.

Twelve watches. Mass channel or role deletion and creation, mass bans and kicks, webhooks created, webhooks deleted or changed, a bot added, a role given dangerous permissions, a member given a dangerous role, and a change to the server name, vanity or owner.
Two speeds. A fast limit per watch, such as 3 channel deletions in 10 seconds, and an optional slow tier over a window up to a day for an account that stays under the fast limit and keeps going. The slow tier is off until you set it.
Alert first, punish if you choose. The default response is an alert and a ping to your mod roles. Opt in to strip the account's roles, quarantine, kick or ban. Moderator roles are deliberately not exempt, because a hijacked moderator account is the case this exists for. The owner and your exempt accounts and roles are never touched.
Dangerous permissions are reverted. A role that gains Administrator, Manage Server, Manage Roles, Manage Channels, Manage Webhooks, Ban, Kick or Mention Everyone goes back to its previous permissions, and a member handed such a role loses it again, whatever the response. The server name and vanity link go back after a takeover attempt. An owner change is reported as something only the new owner can undo.
Mass bans are undone. With a punishing response, every account the attacker banned inside the window is unbanned again with the note Reversed by anti-nuke, up to 200 accounts per incident. Accounts Sentinel or a moderator banned since stay banned. Optionally, everyone the attacker kicked gets a DM with a one-use invite back.
Quarantine hold. An account anti-nuke has punished is held. Anyone who gives it roles back, lifts its quarantine, unbans it or changes a channel permission for it gets the same response and an alert that says why, unless they are exempt. Holds end on release, overturn, appeal approval, /unquarantine or any Sentinel unban.
Panic mode. A punishing response can also lock the whole server down. While that lockdown is on, only the server owner and your exempt admins can run punitive and lock commands or change cases, locks and settings on the dashboard. A compromised moderator cannot lift it, ban the responders or purge the evidence.
Restore the structure. With restore on, deleted channels and roles come back from an hourly snapshot. They come back empty, because Discord cannot restore messages.
On by default where it matters. The setup wizard and the AI setup turn on anti-nuke alerts, escalating timeouts and the message log, and a protection strip on the overview shows what is on and off, with a one-click Turn on recommended protection for servers set up earlier.

Raids, joins and lockdowns

What happens when fifty accounts arrive at once.

Four raid levels. Off. Soft, a 10 second slowmode everywhere. Hard, a 30 second slowmode and every new join quarantined, kicked or banned, with a DM to every admin. Lockdown, nobody can post, verification at its highest, and joins are kicked.
Arms itself. Hard mode switches on at 10 joins a minute by default and switches off after 30 minutes unless you say otherwise. /raidmode and the dashboard buttons override it.
Look-alike detection. Accounts joining within a window that share near-identical names, were created within a day of each other, have no avatar or have ids minted minutes apart are clustered. Five such accounts raise raid mode to Hard, get the raid join action and are reported in one embed. Both windows are settings.
The invite that brought them. Every join is attributed to the invite it came through. The raid report names it and its creator, and when one invite brought the trigger count of look-alikes, Sentinel revokes it and says so. The vanity link is never touched.
Join risk. Account age, no avatar and join velocity, scored the moment someone arrives, with a kick, quarantine or flag at thresholds you set.
Unverified and unwanted bots. A bot without Discord's verified badge can be kicked on join, with the member who added it noted in a case. With the allowlist on, a bot added by anyone but the owner, a listed account or an exempt admin is kicked on sight.
Lockdown, five ways. /lockdown locks this channel, the whole server, one role, new joins or invites. Every lock records what it changed and /unlockdown puts back exactly that. Add minutes and it lifts itself. The overview lists active locks with an Unlock button and can lock the server, joins or invites with a reason and a timer.
The log keeps flowing. A lockdown never locks Sentinel's own log channels, so staff can follow the incident while the server is closed.
A honeypot. A channel humans never post in. Anyone who does is banned, kicked, timed out, quarantined or just flagged, your choice.

Server change logs

Who changed what, from the audit log, in channels of their own.

Seven categories. Members (joins, leaves, kicks and bans, nicknames, roles, timeouts, name and avatar changes), channels, roles, server settings, voice, content (emoji, stickers, invites) and moderation. Each posts to its own channel in its own colour.
Before and after. Every entry shows what changed, when, and who did it from the audit log, with the ids in the footer.
One entry per burst. Several edits to the same channel or role within two seconds become one entry, across every bot replica, paced to one embed a second per channel.
Ignore lists. Channels and roles you name are left out, including name and avatar changes by members holding an ignored role.
Set up in one click. The Logs page and the setup wizard create private #mod-log, #message-log and #server-log visible to your moderator roles and point every category at them.

Cases, appeals and reports

The paper trail.

Numbered cases. Every action, automatic or manual, gets a case with status, staff notes, a timeline, the related reports and any appeal. Export to CSV.
Why this fired. The matched term or pattern, the measured value against the threshold, the join-risk breakdown and the escalation maths, plus the 11 messages around the one that triggered it.
Overturn reverses. Overturning a timeout, ban, quarantine, voice mute or voice deafen undoes it. Blocked if a newer live case of the same type exists.
Appeals. One per case, from /appeal or a web page where members sign in with Discord and see only their own cases. Staff approve or deny with a note.
Reports. /report or right-click a message, Apps, Report Message. One pending report per pair of members, an inbox sorted by risk, and Warn, Timeout, Kick and Ban buttons on the row.
Native actions recorded. Bans, kicks, timeouts, mutes and deafens done in Discord itself become cases too.
A watchlist. /watchlist adds a member by hand, or a risk band adds them for you. A manual entry is never dropped by the automatic release.

Risk, points and repeat offenders

Three numbers that make the rules smarter over time.

A risk score per member. 0 to 100, weighted by case severity and fading over a year, with bands from clean to critical. Crossing a band can add the member to the watchlist, post to the log, ping the mods, or quarantine if you opt in.
Warning points. A ledger per member. Warnings expire after 30 days by default, points can decay daily, and overturned cases do not count.
Longer timeouts for repeat offenders. Each prior timeout in 30 days multiplies the next, 2× by default, up to a ceiling you set.

Discord's own AutoMod

Discord's native rules block a message before it exists, which no bot can do. Sentinel manages them and records what they catch.

Native rules from the dashboard. Create and edit Discord's AutoMod rules here: keyword lists with regex and an allow list, the profanity, sexual content and slurs presets, mention spam and spam, each with block, alert channel and timeout actions and exempt roles and channels.
Every hit in the mod log. Blocks and timeouts by Discord AutoMod post to the mod log with the rule, the matched keyword and an excerpt. On by default. Turn on cases to record each hit against the member, with no warning points.
Rule changes in the server log. An AutoMod rule created, changed or deleted shows in the server log with who did it.

The moderator's day

The everyday tools, done carefully.

Purge. Delete 1 to 100 messages filtered by member, bots, links, attachments or text. Pinned messages are never touched, and the purge is logged.
Softban. Ban and unban in one go to clear a member's recent messages. Its own case type, appealable.
Confirm before it acts. With the setting on, /ban, /softban, /kick, /purge and /timeout show the target and the action with Confirm and Cancel first. Only the moderator who ran the command can answer, and the prompt expires after 30 seconds.
Anonymous moderation. Turn it on and the mod log and punishment DMs say the moderation team acted instead of naming the moderator. Cases still record who acted, for staff.
A default timeout. Set a length in Settings and /timeout works without a duration. A typed duration always wins.
Slowmode and temporary roles. Slowmode by command, logged. Temporary roles from 30 seconds to a year, and never a role with a dangerous permission.
A message log. Deleted and edited messages with before and after, who deleted them, images kept, with your retention and a CSV export.
Ban sync across your servers. Servers owned by the same account can flag, kick or ban on join anyone banned in a sibling, and copy new bans across. A ban network does the same between servers with different owners: a server joins with the network's code and the network owner approves it. Each server picks its own response and whether pushed bans apply to it.
Report by reaction. Set a report emoji and any member can report a message by reacting with it. The reaction is taken back and the report lands in the inbox with the message as evidence. A cooldown stops one member flooding it.
Scan the backlog. Run your rules over channel history, up to 2,000 messages a channel across 25 channels. Report mode lists every hit by rule with a jump link. Delete mode removes the matches and writes one case per author. A scan never kicks, bans or times anyone out.
Roll back a template. Every template install remembers the rules it created. One button removes exactly those rules and nothing else.
Platform-banned accounts. An account banned from YourBot itself is flagged on join by default, or quarantined, kicked or banned. The reason is never shared.
DM templates. Your own wording for the warn, timeout, kick and ban notices, with {server}, {reason}, {duration}, {case} and {moderator}.
AI setup. Describe your server in up to 2,000 characters, pick a strictness, preview the generated rules, apply. It creates the log channels and turns on the recommended protection too.

In the dashboard

Overview The protection strip, open cases, pending appeals and reports, high-risk members, the raid mode controls, the lock panel with active locks and lock buttons, recent activity and the slash command reference.
Cases Filter by status, type, target, moderator and date, bulk resolve, overturn, reopen, archive or note, and export. Each case opens to its evidence panel.
Users The risk roster by band, the watchlist, and each member's history with a score breakdown and a release-from-quarantine button.
Messages The deleted and edited message log with attribution and export.
Logs The seven server-log categories with a channel, a colour and a test button each, the ignore lists, and the Create log channels button.
Reports and Appeals The report inbox with one-click actions, fed by the command, the right-click menu and the report emoji, plus the appeal queue with approve and deny.
Anti-nuke The watch, ping and ignore-bots switches, alert channel, exemptions, the response, hold, panic, the undo and revert switches, restore, each watch's fast and slow limits, the active holds and the recent incidents with what was reverted or undone.
Rules and Templates The rule list with toggles, the editor with condition and action builders, scoping, cooldown, dry run and a Test Rule panel, plus the template library, where every install lists its rules with a Remove rules button.
Discord AutoMod Discord's native rules with their trigger, actions and exemptions, forms for keyword, preset, mention spam and spam rules within Discord's caps, enable, edit and delete, plus the two logging switches.
Scan Pick channels, a depth and report or delete mode, then watch progress, read the hits grouped by rule with jump links or cancel the run.
Analytics and Insights Actions over time, top rules and members, raid events, response times, top moderators and an activity heatmap.
Settings Log channel, quarantine role, mod roles, the moderator command switches (confirm, anonymous, default timeout), the report emoji and its cooldown, the phishing block and allow lists, ban sync and ban networks, warning expiry and decay, evidence retention, dry run for everything, raid thresholds, look-alike detection and its windows, the bot-add allowlist, reaction floods, platform-banned accounts, native action capture, ban sync, the honeypot, DM templates, report handling, the message log, risk bands and escalation, repeat-offender timeouts, and data purge.
Setup The wizard with a readiness check, three protection levels and the log channels, and AI setup from a description of your server. If the bot can no longer see an existing log channel or moderation category, the wizard says so and creates new ones instead of failing.

Configurações

GeneratedAll 32 settings
Moderation log channel Moderation actions, automod hits, purges, slowmode and lockdown changes, raid mode changes, watchlist changes, risk alerts and high-risk report alerts are posted here as cards. Staff notes are not. Empty by default: nothing is then posted in Discord, but cases are still recorded on the dashboard. The setup wizard creates #sentinel-logs and fills this in.
Quarantine role The role /quarantine, auto-quarantine and raid mode give a member. Should remove access to your channels.
Moderator roles Roles treated as staff. Their members can run the moderation commands, are skipped by automod rules, never trigger risk alerts, and are the roles mentioned by high-risk alerts. Empty by default: only Administrators can then run the staff commands.
Warning expiry Days a warning stays active, 30 by default. After that it is marked expired and the member's warning count goes down. The case itself is kept. The Settings page accepts 1 to 365.
DM on warn Sends the member a direct message when they are warned.
DM on timeout Sends the member a direct message when they are timed out.

Runs on its own

Reversals on schedule. Temp bans, timeouts, temp roles, quarantines and timed locks lift themselves. A pending reversal survives a crash and is retried.
Raid mode stands down. After 30 minutes by default, and a panic lockdown with it.
Reverts run within seconds. A dangerous permission change, a renamed server or a changed vanity link is put back as soon as the audit entry arrives.
Warnings age out. Warn cases expire after your window and points can decay daily.
Snapshots every hour. Channel and role structure, kept three deep, when restore is on.
Lists stay fresh. The two phishing feeds are pulled every 6 hours by one process for the whole fleet. Domain-age lookups are cached for 7 days and the invite snapshot follows every invite created or deleted.
Scans run in the background. A queued scan is picked up by the bot that serves your server within 30 seconds, reports progress every 250 messages and stops cleanly when you cancel.
The message log trims itself. Rows past your retention are deleted.

Limites

WhatValorNotas
Timeout1 s to 28 daysDiscord's ceiling. A default timeout can be up to 28 days too.
Ban message deletionup to 7 daysDiscord's ceiling.
Purge1 to 100 messagesPinned messages excluded.
Slowmode0 to 21,600 s
Lockdown timer1 to 1,440 minInvites pause for at most 24 h, Discord's cap.
Temporary role30 s to 1 year
Rule cooldown0 to 3,600 sPer member per rule. Default 10.
Regex in a filter200 charactersAnd at most 6 open-ended quantifiers.
Evidence capture11 messagesAround the trigger, 500 characters each. Reports capture 30 by default, up to 50.
Warning expiry1 to 365 daysDefault 30.
Message log retention1 to 365 daysDefault 30.
Raid auto-trigger10 joins per minuteDefault.
Look-alike cluster5 accounts in 10 minDefaults. Created within 1 h to 7 days of each other, ids minted 30 s to 1 h apart.
Anti-nuke fast window5 to 3,600 sCount 0 to 100. 0 disables a watch.
Anti-nuke slow windowup to 86,400 sOff until a slow count is set.
Anti-nuke undo200 accountsPer incident, unbanned or re-invited.
Bot-add allowlist200 accounts
Confirm prompt30 sThen it cancels itself.
Appeal reason20 to 2,000 charactersOn the web page.
Phishing lists200 domainsBlock and allow each, per server.
Report emoji cooldown10 to 3,600 sPer member. Default 300.
Scan2,000 messagesPer channel, 25 channels per run, one run at a time.
Discord AutoMod rules6 keyword rulesDiscord's caps, plus one spam, one preset and one mention rule.
Ban networks10 per server50 servers per network.

What it watches

Generated

The Discord events Sentinel reacts to while it is enabled. It does not read history from before it was turned on.

Audit log entriesChannel creationsChannel deletionsChannel changesRole deletionsServer setting changesInvites createdInvites deletedMembers joiningMembers leavingMember changes (roles, nickname, timeout)New messagesMessage deletionsMessage editsReactions addedThreads createdProfile changes (name, avatar)Voice joins, leaves and moves

Safety

Protected targets. The server owner and bots can never be timed out, kicked, banned, quarantined, renamed or voice-punished by a rule, a temp role or raid mode. Anti-nuke never acts on the owner or on the accounts and roles you exempt.
Discord's answer is the truth. The bot does not pre-check role hierarchy. If Discord refuses an action, the case says so instead of pretending it worked, and a refused revert is reported in the same embed.
A revert never makes things worse. A permission revert with no recorded previous value is skipped rather than stripping a role bare, and an unban is skipped for anyone Sentinel or a moderator banned again since.
Temp roles cannot escalate. A temp role can never carry Administrator, Manage Server, Manage Roles, Manage Channels, Ban or Kick, and the moderator must outrank it.
Regex cannot hang the bot. User-written patterns are capped and validated before they are accepted.
Evidence outlives Discord. Context messages are cached for 72 hours and attachments up to 8 MB are re-hosted, so a deleted message still has its evidence.
Appeals are private. The appeal page requires a Discord login and shows only that member's own cases in that server.
DMs never block a punishment. A member with DMs closed is still actioned. The notice is best effort.

What it does not do

  • No image or NSFW scanning. Attachment rules look at file names and counts.
  • No message backups. Restore brings channels and roles back empty.
  • No rescue key. The server owner is always exempt and the accounts and roles on the anti-nuke exempt list can lift a panic lockdown. Discord's owner transfer is the recovery path for a lost owner account.
  • Fresh-domain detection is an age check. It flags a link to a domain registered in the last 30 days, not a domain that merely looks like a known site.

Perguntas

Does it read images?

No. It can block file types, floods of attachments and members with no avatar, but it does not look inside an image.

What if a rule fires by mistake?

Start in dry run and read the cases it would have made. Once live, overturn a case and the action is reversed, or let the member appeal.

How do members appeal?

With /appeal and the case number, or on the web appeal page after signing in with Discord. One appeal per case, reviewed by staff with a note.

What does raid mode do?

Soft adds a 10 second slowmode. Hard adds a 30 second slowmode and quarantines new joins. Lockdown stops all posting and kicks new joins. It arms itself at 10 joins a minute, or when look-alike accounts cluster, and stands down after 30 minutes.

Will it stop a hijacked admin account?

Anti-nuke alerts on mass deletions, bans, kicks and dangerous permission changes within seconds, puts the permissions, the server name and the vanity link back, unbans the people the account banned, and can strip its roles, quarantine, kick or ban it. With panic mode on, the server locks down and only the owner and your exempt admins can moderate until it lifts. With restore on, deleted channels and roles come back from the last hourly snapshot.

Who can moderate during a panic lockdown?

The server owner and the accounts and roles on the anti-nuke exempt list. Everyone else, including administrators, is told that panic mode is on until raid mode is turned off.

Do bans carry across my servers?

Yes, two ways. Servers owned by the same account share bans when you opt in. Servers with different owners share them through a ban network: a server joins with the network's code and the network owner approves it. Each server chooses its response on join and whether pushed bans apply to it.

Can I use Discord's own AutoMod too?

Yes. The Discord AutoMod tab creates and edits Discord's native rules, which block a message before it is posted. Every block and timeout by those rules shows in the mod log and can be recorded as a case.

Can I check old messages against a new rule?

Yes. Scan runs your rules over channel history in report mode first, so you see every hit with a jump link, then in delete mode if you want the matches gone. A scan never punishes anyone beyond deleting the message.

AnteriorGroup Finder
YourBot docs Reference tables are generated from the code that is running. Ask in Discord Suggest a correction