Automod

Under /settings → 🛡️ Automod → 🚫 Banned Phrases you maintain two separate lists:

• Exact Phrases — case-insensitive, word-boundary anchored. Inline * wildcards still work — bad*word catches badword, bad123word.
• Wildcard Phrases — every entry matches anywhere inside a word; whitespace in the entry fuzzes to any non-alphanumeric run, so gift card catches gift card, giftcard, gift-card, gift_card, gift.card, MyGiftCard, agiftcardy. Single-word entries like scam catch scam, scammer, iamascammer.

Both lists share the same workflow: case-insensitive matching, zero-width character stripping, an Add modal that accepts comma- or newline-separated batch entries, a Remove-by-select dropdown, and caps of 250 entries per list / 120 characters per entry.

Under /settings → 🛡️ Automod → 🔎 Regex Filter, maintain a list of Python-style regular expressions — each is validated when you add it, so a malformed pattern is rejected rather than saved. Every message and edit is tested against all patterns; the first match deletes the message (same enforcement path as Banned Phrases: delete + optional ladder + notice + audit). Use it for structured spam the phrase lists can't express, e.g. (?i)free\s*nitro. The matched pattern is your own regex, shown in the audit for tuning. Capped at 25 patterns.

Rather than building a slur / profanity list by hand, toggle Voltcord's maintained packs under /settings → 🛡️ Automod → 🧪 Preset Packs — Profanity, Slurs, and NSFW. Any combination can be on at once; matching is substring / wildcard and case-insensitive, so evasions are still caught. A hit takes the same action as Banned Phrases, and the audit shows which pack tripped (not the raw term).

Catches a member posting the same or near-identical message repeatedly. Configure N repeats in Y seconds (default 3 in 30 s), exact or fuzzy matching (fuzzy uses a configurable similarity threshold, default 0.90), and whether the window is per-channel or server-wide, under /settings → 🛡️ Automod → ♻️ Duplicate Filter. Whitespace / case / zero-width differences are normalised so trivial tweaks still count. Tracked in memory (resets on bot restart).

One rule with three independently-togglable sub-detectors under /settings → 🛡️ Automod → 🔠 Text Quality — Caps (flag messages above a percentage of capitals once past a minimum length, default 70% over 10 chars), Emoji (more than a max count, default 10), and Zalgo (combining-character density above a ratio, default 0.30). Cheapest-first; the first to trip wins, and the audit names the detector and the measured value.

Fires when a message carries attachments or stickers. Under /settings → 🛡️ Automod → 🗂️ Attachment Filter, pick a mode — blocklist (block these) or allowlist (allow only these) — then enable any of six MIME categories (Executables / Archives / Images / Video / Audio / Documents), add custom MIME types, set an optional per-file size cap, and list blocked sticker IDs. First violation wins (size → MIME → sticker); the audit names the offending file or sticker and why.

Pick block_all (deny any link), allowlist (only listed domains pass), or blocklist (listed domains are blocked, others pass). Matching is subdomain-aware — example.com also matches sub.example.com — and schemes are optional, so bare domains are still caught. The list is capped at 250 entries.

Pick block_all to deny every Discord invite, or allowlist to let specific servers through. Allow-list entries accept both numeric guild IDs (15–20 digits) and vanity codes — Voltcord resolves an invite to its guild ID at scan time and caches the lookup for 5 minutes. Cap: 100 entries.

Set a per-message mention cap (default 5, range 2–50) and choose which of Users / Roles / @everyone + @here count toward it. On a hit, the offending message is deleted and the author receives a configurable timeout (default 5 min, range 30 s – 7 d). Edits are scanned too.

Configure N messages in Y seconds per user per channel (default 5 messages per 5 s; max-messages 2–50; window 1–60 s). Exceeding the cap deletes the triggering message and applies a configurable timeout (default 5 min, range 30 s – 7 d). Unlike the other message rules, edits do not count — only newly-posted messages.

When a joining member's Discord account is younger than the configured threshold (default 7 days, range 1–365), Voltcord assigns an admin-configured quarantine role. No kick, no ban, no block — the member joins, gets the role, and you control what the quarantine role can see and say through normal Discord channel permissions. New joins are scanned automatically; to catch members who joined before you enabled the gate, run /automod rescan accountage, which quarantines any existing members still under the threshold.

Checks a member's username, global name, and nickname against an admin pattern list — both on join and whenever they change their name or nick. A match takes a configurable action under /settings → 🛡️ Automod → 🔤 Username Filter: Rename (default — resets the nickname to a configurable string such as Filtered User), Kick, or Quarantine (assign a role). Matching is substring-aware, so a banned word buried in a longer name is still caught. The bot skips the server owner and anyone it can't outrank.

Maintains a block list of avatar keys — Discord's own per-image identifier, so no image is downloaded — under /settings → 🛡️ Automod → 🖼️ Avatar Block. On join, and whenever a member changes avatar, a match takes a configurable action: Kick (default), Ban, or Quarantine. Copy a key from a raid audit entry to add it. The server owner and members above the bot are skipped.

Mark specific channels (text / news / voice / forum) as gated, each with its own threshold in whole hours (1–720 = 30 days). When a member posts in a gated channel before they've been in the server long enough, the message is removed and (optionally) a transient in-channel notice explains why. Threads inherit their parent channel's gate automatically. Use this to keep fresh accounts out of announcements, trading, or verified-only channels. Capped at 25 gated channels (Discord select-option limit).

Mark channels where every message must carry at least one real attachment or sticker — text-only messages (including pasted GIF / image links) are removed. Captions are fine: any message with an attachment is valid however long the text. Configure the channel list and optional bypass roles under /settings → 🛡️ Automod → 📎 File-Only Channels; threads inherit their parent channel's rule. Action: delete + optional in-channel notice + audit.

Post a public panel message (title, description, and an ✅ I'm not a bot button) in a channel you pick; new members click the button to receive an admin-configured verified role. Enforcement is passive — you configure Discord channel permissions so @everyone can't speak where the verified role can, and Voltcord only manages the role. The settings wizard walks you through the required pieces (channel perms → role pick → channel pick → post the panel → toggle on) and refuses to enable the rule until everything is in place. The panel's button is a persistent view, so clicks survive bot restarts.

Set a velocity threshold (default 10 joins, range 2–100) and a window (default 30 s, range 5–600 s). When more joins land within the window, Voltcord posts a 🚨 Raid Detected audit embed listing the burst plus up to 5 flagged joiners (mention · account age · join-lag) and persists the event so later anti-raid tooling can act on the same burst.

Pattern Detection (opt-in) layers coordination-based detectors on top of the velocity check, watching the join stream over a separate longer window (default 5 min). Toggle it on under 🚨 Raid Detection → 🛡️ Patterns and enable any combination of three independent detectors:

• Username similarity — flags clusters of joiners whose usernames are pairwise similar above a configurable threshold (default 0.85, range 0.60–0.95). Catches farms like RaidUser01..RaidUser20.
• Account-age cluster — flags clusters of joiners whose Discord accounts were created within the same window (default 60 min) AND in the recent past (default last 7 days). Catches account farms that were mass-created shortly before a raid.
• Avatar collision — flags clusters of joiners sharing the same avatar hash. Cheap and effective: sock puppets that don't customise their avatars get caught (Discord's default avatars count too).

Detectors run cheapest-first (avatar → account-age → username), and the first to trip wins so a single coordinated burst writes one event row with a 🚨 Coordinated Join Pattern Detected audit. Cluster size required to trip is configurable per detector (default 5+, range 3–50).

Auto Raid Mode (opt-in) layers automatic enforcement on top of both velocity AND pattern detection. Toggle it on under 🚨 Raid Detection → 🛡️ Auto Mode and pick a duration (1–1440 min, default 30). On every detection trip — velocity OR pattern — Voltcord engages any combination of three actions you've enabled, then auto-lifts when the duration expires:

• Slowmode — applies a configurable delay (default 30 s, max 6 hours) to every text channel the bot can edit. Only raises per-channel — never reduces a stricter setting an admin already had in place.
• Verification level — raises the server verification level to a configured target (default high). Only raises if the current level is below the target.
• Invite pause — flips Discord's server-wide Pause Invites flag. Only flips it on if invites aren't already paused.

Each engage and lift posts its own audit embed (🚨 Raid Mode Engaged / ✅ Raid Mode Lifted) summarising exactly what changed and what was restored. Re-trips during an active raid mode do not extend the duration — admins know exactly when lockdown ends from the engage time.

Bulk Cleanup Tooling closes the loop. Use /automod raid review (server-administrator only) to open a paginated list of recent events, or click the 🔍 Review button on any 🚨 Raid Detected / 🚨 Coordinated Join Pattern Detected audit embed to drill straight into a single event. The review UI shows every flagged member's current state (in server / banned / left) and offers four bulk actions, with a recommended action highlighted in green based on the detector:

• Ban All — permanent ban on every flagged member still in the server. Skips already-banned members.
• Kick All — kicks every flagged member still present.
• Quarantine All — assigns a configured quarantine role (set under /settings → 🛡️ Automod → Raid Detection; can be different from the Account-Age Gate quarantine role).
• Timeout All — applies a Discord timeout for a configurable duration (default 1 hour, max 28 days).

Every destructive action shows an ephemeral confirmation listing exactly who'll be hit and who'll be skipped, then posts a final audit embed summarising the result. There's also a Mark Resolved (no action) button for cases where a mod has already handled the situation outside the bot.

One-click false-positive cleanup. If the bulk action turns out to be wrong, click ↩️ Reverse Last Action on the same review screen — Voltcord undoes the action against every member it actioned (ban → unban, quarantine → remove role, timeout → clear) and posts an audit embed recording the reversal. The original action stays in the event's history; the reversal is appended on top, so the full audit trail of who-acted-and-who-undid-it survives. Kicks aren't reversible (the kicked member already left), so the button is hidden in that case.

No action is taken on members automatically — Auto Raid Mode tightens the *server*, not its members. Member action only happens when an admin clicks a bulk action button.