JevMade Sign in
← Back to experiments

Apps & data pipelines

AI Werewolf

A social-deduction game where Jev chooses which bot speaks next and separately screens text written by human players.

Bookmark: AI Werewolf Keep this in your collection.
Leave a noteWhat would you try with this? : AI Werewolf

Only you can see your notes.

Source screenshot of AI Werewolf
SOURCE SCREENSHOTFull screenshot ↗

What it does

Human text is screened before reaching chat providers, but character output is not. Jev routes speakers rather than voicing them.

How you can use it

Sketch out the turn rules for your multiplayer game first. List the characters and decide when someone should speak up, such as after an accusation. A developer can use this project as an example to build a referee system that decides who talks next.

Your app can send the current game conversation to a referee AI that picks the best character to reply. This judge only chooses the speaker and checks player messages. It does not write lines, so you still need ordinary chat models to generate dialogue.

Maker-reported (not independently measured by JevMade): Jev is priced at $42 per 1B input tokens with output free (author's constant) · A 2026-09-19 content-screen probe scored "lynch him, string him up" at 0.15, a gore description at 0.88 and a Gemini-refusal framing at 1.97 on the 0–3 risk scale (author-reported)

Primitives
score, choice, noul
Platform
TypeScript
Added
Project created
GitHub stars
20 (snapshot, not live)

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.