ModSentinel

AI-powered Reddit mod tool that auto-detects violations, tracks user reputation scores, sends warning DMs, and delivers weekly reports — saving mod teams 10+ hours per week.

Comment

Inspiration

Reddit moderators are unsung heroes. They're volunteers who spend 10-15 hours a week manually warning users, tracking repeat offenders, writing ban messages, and reviewing violations — all for free. I wanted to build something that gave that time back to them.

The problem I kept seeing across large subreddits was the same: mods had no system. No way to automatically track who had been warned before. No way to know if a user was a first-time offender or a serial rule-breaker. No community health visibility. Everything was manual, inconsistent, and exhausting.

ModSentinel was built to fix that.

What It Does

ModSentinel is an AI-powered moderation tool that runs silently in the background of any subreddit. The moment a post or comment is submitted, ModSentinel scans it for violations — hate speech, harassment, threats, spam, and explicit content. It then takes immediate, graduated action without the mod ever having to lift a finger.

Here's what happens automatically:

  • Strike 1: User gets a personalized warning DM explaining exactly what they did wrong
  • Strike 2: User gets a final warning and a 1-day cooldown ban
  • Strike 3: User is automatically banned for 7 days

Beyond automation, ModSentinel gives mods powerful tools they've never had before:

  • User Reputation Scores: Every user has a dynamic 0-100 score that goes up for good behavior and down for violations. Mods can instantly see if someone is a trusted community member or a high-risk user before making decisions.
  • Community Health Score: A live 0-100 score showing the overall health of the community based on recent violations.
  • Weekly Mod Reports: One click sends a full weekly summary to the mod's inbox — violations by severity, actions taken, health trends.
  • Mod Dashboard: Real-time view of the 5 most recent violations with user names, severity, reasons, and reputation scores.

How I Built It

ModSentinel is built entirely on Devvit — Reddit's developer platform. The backend uses Hono as a lightweight web framework, with Redis for persistent storage of violation logs, strike counts, and reputation scores.

The violation detection engine uses pattern matching across three severity tiers — high (threats, violence), medium (harassment, hate speech), and low (spam). Each tier triggers different automated responses, creating a graduated enforcement system that mirrors how good human moderators actually think.

The reputation system was the most innovative piece. Rather than just punishing bad behavior, I wanted to reward good behavior too. Every clean post or comment earns +2 reputation points. Violations deduct 5-30 points depending on severity. This creates a complete picture of a user's community standing over time.

Challenges I Ran Into

The biggest challenge was working with Devvit's newer web server architecture, which uses Hono routes rather than the classic Devvit Blocks approach. Getting Redis to work correctly through the server context required deep diving into how the template's internal modules expose their clients — specifically discovering that redis and reddit could be imported directly from @devvit/web/server, the same pattern used in the template's own core utilities.

Another challenge was designing the graduated response system to be useful without being annoying. Too aggressive and mods would uninstall it. Too passive and it wouldn't save any time. The 3-strike system with escalating consequences was the right balance.

What I Learned

Building ModSentinel taught me how much unmet need exists in the mod tooling space. Reddit has millions of volunteer moderators and almost no automated infrastructure to support them. Every single feature I built — the strike tracking, the reputation scores, the health dashboard — was solving a real pain point that mods have been living with for years.

I also learned that the best automation doesn't replace human judgment. It informs it. ModSentinel doesn't ban users without giving them chances to course correct. It doesn't hide information from mods — it surfaces it. The goal was always to make mods more powerful, not to remove them from the equation.

What's Next

  • Custom rule configuration so mods can define their own violation patterns
  • Reputation leaderboard showing top community contributors
  • Peak toxicity hours analysis showing mods when to be most active
  • Integration with Reddit's native AutoModerator for hybrid enforcement

Built With

Share this project:

Updates