Server Assistant🤖 AutoMod
AutoMod is Server Assistant’s automatic message filter. It watches your channels and quietly acts on scams, slurs, spam and dodgy links — so your team isn’t firefighting every bad message by hand. You decide how strict it is, and nothing acts that you haven’t switched on.
When you first set the bot up, AutoMod is seeded from your community type: the setup wizard presets the filter packs, punishment ladder and anti-raid sensitivity to match a Gaming, Creator, Anime, Crypto, Education, NSFW or General community. This page is where you fine-tune all of it after the fact.
Who can configure: an admin role. Some panels — AI Review in particular — need PREMIUM.
The /automod command
Open the AutoMod dashboard — one panel with every filter, list and threshold. Admins only; everything is done with buttons.
| Argument | Type | Notes |
|---|---|---|
| No arguments — the dashboard is fully interactive. | ||
The top of the dashboard is a live summary — whether AutoMod is on, which packs are active, your custom filter and link-block counts, the mention limit, anti-spam status and AI review. Every button below opens one of the panels you can step through next.
Step through every panel
Use Next › and ‹ Back to walk each AutoMod panel, including the forms that open from them.
Filter Packs. Tick the ready-made packs you want — several can run at once. scams blocks the usual Free-Nitro and gift-card scams; slurs_l1 catches slurs while leaving everyday swearing alone; slurs_l2 adds targeted harassment; profanity_l3 filters all casual swearing (warned as high false-positive); ip_grabbers blocks IP-logging links; and lexicon is a broad community word list, refreshed daily.
AI Review PREMIUM. A second opinion for the grey-area hits. When a word filter or spam check is unsure, the bot asks the AI for a confidence score: if it lands at or above your threshold it acts automatically; below that, the message stays up and your staff get a Delete & warn / Allow prompt. Only borderline messages are ever sent for review, so token use stays low.
The threshold form. Opens from 🎚️ Set threshold. Enter a number from 0 to 100. A higher number means the AI has to be more confident before it acts on its own — anything below the line is handed to your staff to decide.
Self-training. With learning on, the bot remembers the content of messages your team deletes or flags via Message Report, then clusters the patterns and — when you press ✨ Suggest rules now — proposes custom filters tailored to your server. Only message text is kept (locally, for 30 days, no author identities), and nothing is enforced until you accept it. 🗑️ Clear learning log wipes the collected samples.
Reviewing proposals. Pressing ✨ Suggest rules now walks you through each proposed rule one at a time — the reason for it, the pattern it would use, and real examples it would have caught. ✅ Accept adds it to your custom filters (live immediately, just like any custom filter); ⏭️ Skip discards it. At the end you get a tally of how many were added and skipped.
Custom Filters. Your own word/phrase patterns, on top of the curated packs. ➕ Add filter opens a form for a regular-expression pattern (up to 200 characters). The bot checks it for you — invalid patterns, or ones that could hang on malicious input, are rejected with an explanation. The dropdown removes any you no longer want.
The custom filter form. Enter one pattern. If it isn't valid, or looks like it could be slow on crafted input, the bot tells you why and asks you to rewrite it — so a stray pattern can never wedge moderation.
Link Blocks. A blocklist of domains you never want posted, on top of the built-in IP-grabber and scam lists. ➕ Add link block takes a domain (e.g. badsite.com) — paste a full URL and the bot strips it down to the domain for you. The dropdown removes entries.
Punishment Ladder. Decides what happens as a member racks up warnings. Each line is count:action[:minutes] — for example 3:timeout:10 times someone out for 10 minutes at 3 warnings, then 5:timeout:60, then 10:ban. Actions are timeout, kick, ban or warn; the minutes are optional. The default ladder is exactly the one shown.
Spam Settings. Two quick controls: the maximum mentions allowed in a single message (anything over the limit is treated as mention spam), and an anti-spam toggle (yes/no) that catches the same message repeated over and over.
How community-type presets seed AutoMod
You never start from a blank slate. When you run /setup
and pick your community type, the bot seeds AutoMod, the punishment ladder
and anti-raid to match — and even pre-guesses the type from your server’s name and
description (defaulting to General if it’s unsure, rather than guessing wrong).
| Community type | Filter packs | Ladder | Anti-raid |
|---|---|---|---|
| 🌐 General / mixed | Scams + IP-grabbers | Standard | Normal |
| 🎮 Gaming | Scams + IP-grabbers + slurs | Standard | Normal |
| 🎨 Creator / Streamer | Scams + IP-grabbers + slurs | Standard | Strict |
| 🌸 Anime / Fandom | Scams + IP-grabbers + slurs | Gentle | Normal |
| 💰 Crypto / Trading | Scams + IP-grabbers + slurs | Strict | Strict |
| 📚 Education / Study | Scams + IP-grabbers | Gentle | Normal |
| 🔞 NSFW / 18+ | Scams + IP-grabbers + slurs | Standard | Strict |
| 🛠️ Set it up myself | (nothing — tune it here) | — | — |
The three ladders are Gentle (5 → 10 → 20 warnings), Standard (3 → 5 → 10) and Strict (2 → 3 → 5), each ending in a ban. Whatever the preset sets, you can override any of it from this dashboard at any time.
See also
- The
/setupwizard — where the community-type preset is chosen - Settings hub & configuration — Quick Presets and every other dashboard
- Wiki home — every command and flow