Babblebox, explained clearly.

Start with one lane, not every command.

Party servers usually start with /play and /daily. Retention-focused communities often add /drops status and scheduled Question Drops next. Smaller communities often start with /watch, /later, /remind, or /bremind. Identity-heavy servers tend to start with /buddy and /profile.

Showable where it should be. Quiet where it should stay personal.

Babblebox separates surfaces people want to share from surfaces that work better as personal tools. That keeps the bot feeling native in Discord instead of turning every interaction into a private panel or every reminder into channel clutter.

Staff-blind anonymous posting with calm, bounded safety.

Confessions are optional in Babblebox and do nothing until admins enable and configure them. This staff-blind lane stays quiet until it is deliberately opened. Once live, /confess opens a private composer, /confess manage, /confess appeal, /confess vote, /confess report, /confess reply-to-user, /confess notifications, and /confess about mirror the private member utility flows, and server staff still moderate by confession ID, vote ID, and case ID only while Babblebox still enforces safety internally without exposing authors or voters.

Short multiplayer energy, built around live rooms.

/play acts as the shared lobby for the live-party lane instead of forcing people to remember several different entry points.

Three quick daily booths plus a guild-first Question Drops lane.

Daily Arcade gives people something to do even when there is no full group ready for a live room. Question Drops add lightweight scheduled prompts, mastery roles, scholar ranks, guild knowledge boards, and private member role controls while keeping channel noise low. Both lanes stay compact, public-friendly, and spoiler-aware.

One identity layer, not a bloated economy.

Babblebox's identity surfaces are meant to feel worth showing. They collect your ongoing flavor, streaks, achievements, party-game wins, Question Drops progress, mastery flavor, and scholar rank into a few readable surfaces rather than exploding into a giant inventory game.

Quiet tools for real server life.

Utilities are where Babblebox earns its keep on ordinary days. They reduce missed context and small task friction without filling channels with bot chatter.

Optional plans, clearer onboarding, and no fake premium mystery.

Babblebox does not paywall core local Shield protection, core Confessions privacy, or the current free utility baseline. The separate Top.gg Vote Bonus lane is temporary, utility-only, not a premium tier, and intentionally smaller than paid plans so it never blurs into Babblebox Plus. It stays off until an operator explicitly sets TOPGG_ENABLED=true. Supporter is the paid support tier with Inevitable Friendship benefits but no Babblebox limit increase, Babblebox Plus raises personal Watch, reminder, and AFK limits, and Babblebox Guild Pro raises bounded server caps, enables guarded setup sync between servers the same owner controls with safe partial imports and skipped-item warnings, can make the higher Shield AI model tiers available when owner policy and provider/runtime readiness allow second-pass review, can make optional Question Drops AI celebrations available when celebration policy and provider/runtime readiness allow live copy, and raises the safe Confessions image ceiling. Premium state stays in a separate entitlement layer, and downgrade preserves saved state instead of deleting it while saved config stays preserved, future expansion stays blocked until you trim it or premium returns, and extra runtime headroom simply pauses until the saved state is reduced or premium returns.

Compact safety for moderators, with honest confidence.

Shield live moderation is optional and starts from local rules. It can look for likely privacy leaks, invite or promo abuse, a first-class Anti-Spam lane grounded in explicit message-rate and near-duplicate thresholds, optional off-by-default emote, capitals, and low-value chatter controls, a separate GIF Flood / Media Pressure lane for repeated low-text GIF disruption, malicious or scam links, no-link DM-lure scams that route nitro, crypto, prize, or cash bait into DMs or off-platform contact, adult-domain intel, an optional conservative solicitation / DM-ad detector inside the Adult Links + Solicitation pack, and a separate Severe Harm / Hate pack for sexual-exploitation solicitation, explicit self-harm encouragement, eliminationist or dehumanizing hate, and severe slur abuse with quote, reporting, moderation, and rules-context suppressors. It can also rescan meaningful edits and webhook or community-post style deliveries in scope, but it is designed to stay conservative when the signal is weak, especially for ordinary bot or webhook posts that only carry medium or soft scam heuristics. Attachment filenames stay metadata-only for link repetition unless a real link was actually posted. Shield also has a separate Trusted Links Only live-message mode for servers that want the built-in trusted pack plus bounded domain or invite allowlists only, and that mode stays separate from Confessions link mode. /shield-links trusted-view now exposes the built-in trusted families, direct domains, and any local built-in trust disables. Specific adult-topic channels can relax only the optional solicitation detector without weakening adult-domain, scam, malicious-link, or trusted-brand impersonation protections. The premium configuration path is /shield panel: pick a pack, then use Actions, Options, or Exemptions so unrelated spam or GIF controls never spill into Severe or Hate work, while each pack can inherit the global timeout or keep a dedicated timeout profile. The Options editor now switches between focused anti-spam or GIF lanes instead of dumping every threshold into one screen. Safe mainstream domains are bypassed before link suspicion, local malicious-link checks include a feed of about 200k known malicious domains, and phase 1 does not depend on any outside reputation provider. Trusted-brand impersonation and lookalike phishing hosts are also hard-blocked locally. Unknown risky links only escalate after combined local scam evidence, hard local intel, dangerous file targets, or fresh-campaign corroboration, and /shield test plus Shield logs show concise local link-decision explanations without exposing raw scoring internals. The first live enable applies a recommended non-AI baseline, while Anti-Spam stays local-first with short-lived windows plus bounded corroboration so healthy active chat stays safer while real spam can still log, delete, timeout, or combine delete + timeout when the configured evidence is strong enough. The main Anti-Spam and GIF lanes can now be turned off individually, optional emote, capitals, and low-value chatter lanes stay off until an admin enables them, moderators are exempt from Anti-Spam by default unless an admin deliberately chooses delete-only or full enforcement, repeated same-member spam alerts group into one evolving incident log instead of spamming the log channel, and pack-specific exemptions let one channel, role, or member bypass one pack without weakening the rest. Shield log buttons stay reversible-first: moderators can open or delete the message when relevant, remove a Shield timeout when relevant, and use the support link without kick, ban, or timeout-punishment shortcuts; confirmed false-positive repair appears only after Shield or the alert delete button changed a message or timeout. Destructive Shield actions can keep a short-lived recovery copy so confirmed false-positive repair can DM the member an apology plus their message text, while Shield still avoids a long-term message archive. The separate GIF Flood / Media Pressure pack keeps GIF disruption legible instead of burying it under generic spam, keeps grouped incidents per offender and incident window for one-user floods, splits one-user GIF floods from collective channel pressure, groups collective pressure per channel, exposes one-member rate, low-text repeat gate, same-asset repeat, true channel-streak, and effective GIF-pressure settings with lightweight meaningful-text weighting, deletes the matched burst for one-user abuse, removes the exact live GIF streak for true consecutive streak floods, trims only the newest contributing GIF posts from the active pressure slice for channel-pressure matches after lightweight meaningful-text weighting, can combine shared cleanup with one member's personal enforcement in mixed incidents, never touches healthy text, and now exposes tighter low-end controls for stricter rooms while keeping the panel usable. The direct moderation suite adds /ban add, /ban remove, /kick, /timeout add, /timeout remove, /quarantine add, /quarantine remove, /warn, /warns, /cases view, /modcases, /notes add, /purge, /sanitize, /slowmode, /verify, and /moderation warnings. Purge is preview-first with moderator caps and log requirements, warning forgetting keeps durable history while only recent active warnings count, and cases/notes stay compact instead of becoming a deleted-message archive. /antinuke status, /antinuke panel, /antinuke permits panel, /antinuke permits add, /antinuke incident, /antinuke release, /antinuke exempt, /antinuke pause, and /antinuke test cover the default-on Anti-Nuke Guard with compact minute/hour windows, scoped permits, command locks, bounded audit-log checks, safe quarantine, quarantine-hold, and owner-reviewed release without deleted-message bodies or server backups. /raidguard panel, /raidguard status, /raidguard test, /raidguard recent, and /raidguard release cover restart-safe join-raid protection that can alert on pure volume but only contain suspicious joiners when corroborating signals make the campaign credible; it uses quarantine or timeout, trusted-role/member/bot suppressors, event mode, compact incidents, false-positive resistant recovery, and Discord Security Actions for bounded invite and member-DM pauses when permissions allow. The linked Admin lane focuses on one direct emergency lock lane plus delegated Role Tools, follow-up, exclusions, logs, and truthful permission diagnostics. /lock add, /lock channel, /lock remove, /lock status, and /lock update give moderators who can manage channels or messages, timeout, kick, or ban members, plus admins one calm direct moderation lane by default, /lock settings lets admins tune the default notice and optionally make that lane admin-only, /timeout remove gives moderators a direct timeout-clear lane, /giverole and /removerole give configured operators a logged single-member role lane, and /admin permissions explains which missing bot permissions degrade locks, timeout removal, Shield actions, or role-based helpers. Separate from that live lane, Shield also supplies private bounded feature-surface checks for Confessions, AFK reasons, reminder text plus public reminder delivery, Watch keywords, and bump reminder copy using configured local Shield text/link checks where guild context exists. Baseline privacy, adult, and severe checks still protect supported fields without guild context. Those private feature checks do not create mod-log alerts, call Shield AI, or run multi-message GIF/spam pressure.

Shield image scanning stays local-first and is enabled by default when Shield is enabled unless admins explicitly turn it off. Babblebox can scan eligible Discord-hosted image attachments for visible scam or phishing text with local OCR, optional local QR decoding, strict size and queue limits, hash/fingerprint dedupe, and confidence-based actions. OCR scam scoring weighs reward bait, trusted-brand bait, account/security action, dangerous secret requests, QR login language, crypto/wallet language, suspicious domains, and known-bad image evidence together, so fake Nitro verification lures are treated as strong image phishing while generic words like free, verify, account, QR, or support stay weak by themselves. Dry runs and alerts show confidence separately from the configured Shield action, so Log only can mean policy, not weak evidence. For flagged image alerts, the original flagged image upload may be attached to the configured Shield log channel for moderator review; raw images are not persisted in Babblebox storage. AI never receives images, image bytes, base64, screenshots, raw QR payloads, or raw image URLs; at most, optional Shield AI review sees sanitized OCR text plus compact metadata after local Shield has already flagged the case.

Dangerous-file scanning is opt-in under the Scam pack Options lane. It is a compact extra layer, not a full antivirus: Babblebox scores risky executable/script/installer extensions, spoofed double extensions, macro-capable files, archive lures, and operator-owned known-bad SHA256s. Optional reputation checks are hash-only, no uploads, and send only SHA256 values to configured providers. Babblebox does not execute, decompress, inspect macros, upload files to VirusTotal or MalwareBazaar, or attach suspicious malware files to logs.

Shield Profile Monitoring is opt-in and checks Discord API-available identity fields: username, global display name, server nickname, and active display name. It can catch blank or invisible usernames, hard scam/adult/impersonation links, and high-confidence scam or severe-harm profile text, then use a state-preserving quarantine role with stored role IDs for release. Weaker combined signals enter bounded Profile Watch instead of quarantine. It does not promise bio, About Me, custom status, or profile-picture scanning.

Common questions