Early development — not yet 24/7

Moderation that watches so you don't have to

Soter moderates mostly on its own: every message is scanned for hate speech and spam, and repeat violators are timed out automatically.

Example: a member posts hate speech, Soter deletes it, and the member gets a DM explaining why.
Powered by Jev

Built for decisions, not conversation

Jev is a model by TypeSafe AI made specifically for quick, structured decisions. Instead of generating text, it returns typed answers with probabilities.

Soter asks it two things about every message — is it hate speech, and does it look like spam — then acts on the answers: clear cases are removed, borderline ones go to your mods.

Learn more about Jev
Example: Jev judges a spam message and Soter removes it.
new_user_4821joined 2 minutes ago

@everyone free nitro for the first 50 people, dm me to claim

Jev
is_hate_speech
0.01
spam_level
high_spam · 0.96
Removed and logged to #mod-log

Hands-off moderation

Set it up once, then let Soter handle the day-to-day.

Hate speech detection
Every message is scanned automatically and flagged the moment it crosses the line.
Spam filtering
Repeated and low-effort spam gets caught before it floods your channels.
Automatic timeouts
Repeat violators are timed out on their own — no mod has to be watching.
Channel exemptions
Exempt specific channels from moderation with a simple settings command.
Configurable thresholds
Tune violation thresholds and timeout duration to fit how strict you want to be.
Mod action logging
Every automated action is logged to a channel you choose, so nothing happens silently.

Built in the open

Soter is open source, written in TypeScript on Bun, with Jev (via OpenRouter) doing the message analysis. MIT licensed.

Put Soter on watch

Add the bot to your server in a couple of clicks and let it handle the day-to-day moderation.