Independent, tunable filters for spam, links, invites, mentions, and words — plus AI-powered scam and phishing detection that reads intent, not just keywords.
Discord's own native AutoMod is keyword- and pattern-based — genuinely useful, but blind to anything that doesn't trip a listed word or phrase. A scam link on a freshly registered lookalike domain, a creator-impersonation giveaway phrased just carefully enough to dodge a word filter, or a phishing screenshot posted as an image instead of text all slip straight through a keyword-only system. That gap is exactly where a dedicated AutoMod layer earns its keep — evaluating the message as a whole, not just scanning it for a blocklist match.
Every rule type in Novrix's AutoMod is independently enabled, disabled, and tuned from the dashboard — set your own thresholds, choose the action (delete, warn, timeout, kick), and add bypass roles for staff or bots that shouldn't be affected.
| Rule | What It Detects | Action Options |
|---|---|---|
| Word Filter | Blocked words/phrases (supports wildcards) | Delete, warn, timeout, kick |
| Spam Detection | X identical messages in Y seconds from one user | Delete, timeout (auto-escalating) |
| Caps Filter | Messages over X% uppercase | Delete, warn |
| Mention Spam | Messages with more than X @mentions | Delete, timeout |
| Link Filter | All links, or links not on your allowlist | Delete, warn |
| Invite Filter | Discord invite links to other servers | Delete, warn |
| Emoji Spam | Messages with more than X emoji | Delete, warn |
| New Account | Messages from accounts younger than X days | Delete, quarantine role |
Novrix analyzes links, images, and text for fake-giveaway language, phishing domains, and impersonation attempts — including scam content inside image attachments and QR codes, not just plain text.
Exempt staff, verified members, or specific bots from individual rules without turning off the filter for everyone else.
Watch AutoMod catches happen in real time from the dashboard, so you can validate a new rule before escalating its punishment.
Start with delete-only, then move to warn or timeout once you've confirmed a rule has no false positives on your server.
AutoMod acts on every single message instantly, without waiting for a moderator to be online. That speed matters — a scam link that sits in a channel for even thirty seconds can be clicked by dozens of members in an active server. AutoMod's job is to make sure the obvious, high-volume threats (spam, scam links, banned words) never get that thirty seconds in the first place, so your human moderators can spend their time on judgment calls instead of copy-pasted garbage.
AutoMod also pairs directly with Anti-Raid — while Anti-Raid watches how fast accounts are joining, AutoMod watches what those accounts do once they're in, catching the message-spam wave that often follows a raid attempt.
Novrix's full AutoMod suite — spam, link, word, and AI scam filters — is free, forever. Run /setup after inviting to get sensible defaults instantly.