Docent screens every message and image with a two-layer AI, then hands the call to your moderators. It flags — your team decides. No automated bans, ever.
Set up in minutes from the dashboard — starts in observation mode, so nothing changes for your members until you're ready.
Three stages between a risky message and a moderator's decision — most messages pass the first one and are never stored.
Every message and image runs through a fast pre-filter; only borderline content escalates to a context-aware LLM that knows your server's rules. Obfuscated text is normalized first, and edits are re-scanned.
Borderline posts can be held briefly before anyone is flagged — the author gets a private nudge and a chance to rewrite or withdraw the message on their own.
Flags arrive as cards in your mod channel with full context. Moderators approve or reject each one — and every decision teaches the AI what your community actually considers a violation.
Soft moderation: firm enough to protect your members, gentle enough to keep them.
A fast pre-filter clears the obvious cases; a context-aware LLM reviews the nuanced ones against your community's own rules.
Instead of instant punishment, borderline posters get a private moment to edit or withdraw — most incidents end there, with no mod involved.
Every moderator decision becomes precedent. Repeat violations resolve instantly, and the AI's judgment converges on your team's.
Members can appeal a decision right from their DM. A moderator upholds or overturns it — overturns reverse strikes automatically.
Members flag concerning posts with an emoji reaction. Enough independent flags send the post to your mod channel for review.
A web dashboard for setup, flagged-message review, strikes, and settings — with an audit log of every action, human or AI.
No. The AI flags content for human review — it never bans anyone on its own, and your moderators approve or reject each flagged item. New installs also start in observation mode: the AI logs what it would have flagged so you can tune it, while your members notice nothing.
For the large majority of messages — anything the AI doesn't flag — the content is analyzed in the moment and discarded, never stored. Flagged content, moderator decisions, and community-confirmed precedents are retained to power the dashboard, audit trail, and learning. Details live in the Privacy Policy.
Docent is free during early access. Paid tiers (billed through Stripe) arrive later — existing communities get clear notice before anything changes. See the pricing page for the planned tier structure.
Add the bot, log into the dashboard with Discord, and a three-step wizard walks you through choosing a mod channel and your rules. You start in observation mode and flip on enforcement when the AI's verdicts look right.
During early access a member-count limit may apply to newly joining servers, so moderation quality stays high while we scale. If your server is over the limit the bot says so politely and leaves without setting anything up or keeping any data — and larger communities can get in touch to be pre-approved.
Add Docent to your server and see what it would flag — before it changes anything.
Add to Server