Inspiration

Moderators spend a lot of time reviewing repetitive risky posts such as scams, spam, crypto promotions, harassment, misleading claims, and privacy-related content. We wanted to build a tool that reduces this manual effort and gives moderators a faster, clearer, evidence-based way to make decisions directly inside Reddit.

What it does

ModPulse AI is a Reddit Devvit moderation triage tool that analyzes post content and generates a structured moderation report. It provides a risk score, risk level, detected violation categories, evidence summary, suggested moderator actions, user-facing removal message, and estimated moderation time saved.

How we built it

We built ModPulse AI using Devvit, TypeScript, Hono, and Node.js. The app integrates into Reddit’s post moderation menu, allowing moderators to open the analyzer from a post, enter the post title/body, and generate a full moderation report. The system uses an AI-ready architecture with OpenAI support and a reliable rule-based fallback engine for consistent risk detection.

Challenges we ran into

One major challenge was working within Devvit’s current form and UI limitations while still creating a useful moderator experience. We also had to handle API restrictions, safe fallback scoring, report formatting, and reliable local playtesting. Another challenge was designing the analysis output to be simple enough for moderators to trust, but detailed enough to support real moderation decisions.

Accomplishments that we're proud of

We are proud that ModPulse AI works directly inside Reddit’s moderation workflow and produces clear, actionable reports instead of only showing a simple alert. It can identify multiple risk types, recommend actions, estimate time saved, and generate removal messages that moderators can immediately use. We also built it with both AI integration and fallback reliability, making it practical even when external AI access is unavailable

What we learned

We learned how to build and test a Devvit moderation tool, configure Reddit menu actions, handle form-based workflows, and design moderator-first user experiences. We also learned the importance of explainability in AI moderation: moderators do not just need a score, they need evidence, categories, and clear next steps.

What's next for ModPulse AI

Next, we plan to add automatic post content fetching, a full dashboard view, downloadable PDF reports, subreddit-specific rule configuration, mod queue analytics, repeated-offender tracking, and deeper AI-based policy reasoning. The long-term goal is to make ModPulse AI a complete moderation intelligence layer that helps communities reduce workload, improve safety, and make faster, more consistent decisions.

Built With

Share this project:

Updates