Incident Response

Join the dots between Automod, raids, impersonation, scams and nuke attempts, score each incident by severity, and respond automatically, up to locking down the whole server.

  • Advanced
  • 12 min read
  • 5 steps
  • Updated 13 November 2026

By the end, you'll have

  • Signals from your other protection plugins grouped into incidents with a severity
  • Automatic responses for each severity, from timeouts to a full lockdown
  • Detection of nuke attempts, like mass channel or role deletions

Before you start

  • At least one signal source set up, such as Automod.
  • Dreamliner with View Audit Log for nuke detection, and Manage Channels and Manage Roles for lockdowns.
  • Manage Server, so you can open the dashboard.

One Automod hit isn't an emergency. Ten from the same person in a few minutes, plus a scam catch and an impersonation alert, might be. Incident Response watches your other plugins, merges related signals into one incident and raises its severity as evidence builds up.

One incident

Related signals within 30 minutes merge together.

Four severities

Low, Medium, High and Critical, from a risk score.

Automatic response

Choose actions for each severity, including lockdowns.

Step 1 · about 2 min

Choose signal sources and alerts

Decide what feeds Incident Response and where it posts.

  1. On the dashboard, open Moderation and click Incident Response. Make sure it's enabled.
  2. Under Signal sources, choose which detectors count: Automod, Raid detection, Impersonation Detection, Scam Protect and Server-nuke detection.
  3. Under Alerts, pick a Log channel (it falls back to your moderation log) and the Correlation window, how long signals keep merging into the same incident (30 minutes by default).
Incident Response Signal sources: Automod, Raid detection, Impersonation Detection, Scam Protect and Server-nuke detection
Choose which detectors feed Incident Response.

Step 2 · about 1 min

Set the risk thresholds

Decide how much evidence each severity needs.

Risk thresholds: Medium 6, High 14, Critical 26
The default thresholds.
SeverityRisk scoreMeaning
LowBelow MediumA single minor signal. Logged, no action by default.
Medium6A few signals, or one moderate one. Worth a look.
High14Strong or combined evidence.
Critical26Severe or compounding evidence, like a nuke pattern.

Step 3 · about 3 min

Choose actions for each severity

Respond automatically when things get serious.

For each severity you can add actions and Ping roles, which are pinged the first time an incident reaches that level.

ActionWhat it does
TimeoutTimes out the person involved, for the Duration (minutes) you set.
KickRemoves them from the server.
SoftbanRemoves them and deletes their recent messages (Delete messages (days)).
BanBans them.
Lock the channelStops members sending messages in the channel involved.
Lock the whole serverStops members sending messages, adding reactions and creating threads everywhere.

Step 4 · about 2 min

Tune nuke detection

Catch someone trying to wreck your server.

Server-nuke detection watches the audit log for bursts of destructive actions by the same account within a short window (about 30 to 60 seconds). Set how many of each counts as a nuke:

Server-nuke detection counts: channel deletions, role deletions, bans, kicks, webhook creations and new-account admin grant
Each count is measured within a short rolling window.
SettingWhat it counts
Channel deletionsChannels deleted
Role deletionsRoles deleted
BansMembers banned
KicksMembers kicked
Webhook creationsWebhooks created
New-account admin grantAdministrator given to an account created within this many hours (0 = off)

Step 5 · about 2 min

Review and close incidents

Keep track of what happened.

Open Moderation → Incident Response → Incidents to see every incident with its severity, risk score, status and first and last seen times. Click one for its Signal timeline and Entity graph, then Close incident when you're done.

CommandWhat it does
/incident listShows open incidents.
/incident viewShows one incident in detail.
/incident resolveCloses it as dealt with.
/incident dismissCloses it as a false alarm.
/incident unlockLifts a lockdown early.

Moderator can use these by default. Incidents don't close on their own, so close them once handled.

You're set up

What you just did

  • Chose signal sources, alerts and thresholds
  • Set up automatic responses for each severity
  • Tuned nuke detection and learned to manage incidents

Pair it with Raid Defense Mesh so partner servers hear about raids too.

Common questions

Can I start a lockdown by hand?

Not with a command yet. Lockdowns come from incident actions; use Discord's own permissions for a manual lockdown.

Nuke detection isn't working.

Dreamliner needs View Audit Log to see who deleted what.

Read next