GuardBot investigates harmful messages through a structured conversation, weighs both sides, then decides. No snap judgements. No bans for context your moderators missed.
Every moderation decision is explained, logged, and reviewable. No black box.
Before acting, GuardBot replies to the flagged message, hears from the accused, then asks the victim whether they felt targeted. The decision reflects both sides.
Every discord.gg link is evaluated for cam-site bait, free Nitro scams, and hacked account patterns. Jokes and quotations are distinguished from real threats using confidence scoring.
Images are described by a vision model and fed into the same moderation pipeline as text. CSAM and doxxing images trigger immediate action regardless of the accompanying text.
Every decision is stored with the original message, AI reasoning, and the full conversation thread. Filter, search, and export to CSV directly from the dashboard.
Set individual channels to strict, normal, relaxed, or off. Add a custom note that gets sent to the AI as context. Banter channels get treated differently from help channels.
Explicit threats with personal information, targeted slurs, CSAM, and doxxing are handled immediately without conversation. The moderator is pinged with full context.
Every message that warrants attention goes through each stage in order.
Free. Runs on every message. If content is not flagged, the pipeline stops immediately. No credits consumed. No AI cost. Most messages never leave Stage 1.
The flagged message, last 20 channel messages, user history, and your custom context prompt are sent to GPT-5.4-nano. It determines whether the flag is a true positive, identifies the victim, and classifies severity. False positives are logged but no credits are charged.
The bot addresses the accused, hears their side, then contacts the victim. The final decision is made by the same model with full context. Every exchange is logged. The moderator is pinged when the bot is uncertain rather than acting unilaterally.
Pay monthly based on server size. LTC payments receive a 50 cent discount on any fixed plan.
Credits never expire. Minimum bundle: 500 credits. LTC discount applies to fixed plans only, not credit bundles.
Add GuardBot to your server and set it up in minutes. No moderation happens until your plan is active.
Add GuardBot