Nudger Bee

The Inspiration

The inspiration for Nudger Bee came from looking at the overwhelming operational math that Community Health Workers (CHWs) face. When a single CHW is managing a group of 80 or more chronic-illness patients, they are completely consumed by day-to-day operations - running the group, distributing updates, and answering immediate crises.

They simply do not have the time for individual focus. In a crowd that large, it is functionally impossible to track everyone. As a result, when a patient goes quiet, they often stay quiet. Furthermore, walking into a support group of 80 strangers is incredibly intimidating for a new member. I built Nudger Bee to bridge this gap: to automate the social infrastructure so members don't get lost in the crowd, and CHWs don't have to play manual matchmaker.

How I built it

I built Nudger Bee as an advanced Slack agent designed to handle the structural networking and background coordination a busy human leader doesn't have the hours to facilitate. At its architectural core, Nudger Bee operates as a Model Context Protocol (MCP) client. It connects to an in-repository MCP "program server" that acts as the single source of truth for the community's session calendar and follow-up scheduling. When a Hivemate asks, "What sessions are coming up?", the agent calls the MCP tool live, fetches the real-time data, and executes a visible task_update step right on the screen.

The agent's user-facing workflow is split into four distinct automation pipelines:

1. Multilingual Onboarding & Peer Matching: The agent greets new members ("Hivemates") via a warm, conversational DM in their own native language. To break through the "crowd paralysis" of an 80-person group, it pairs them with a real peer in a similar situation, makes a bee-brokered warm introduction.
2. The Engagement Loop: It runs daily, personality-filled check-in nudges and celebrates consistency by gamifying accountability streaks ("Honey" 🍯) to keep members feeling involved.
3. The Silent-User Pipeline: The agent continuously monitors background engagement thresholds. If a member's inactivity time delta passes a critical threshold, Nudger Bee surfaces their name to their Hive Keeper (the CHW) in a single, grouped daily digest so no one slips through the cracks.
4. The Clinical Fire-Wall: Nudger Bee stays strictly in its lane as a coordinator, never a clinician. I engineered an intent-classification layer so that any symptom report or medical question is instantly refused across all languages. The agent automatically points the user to real medical care while simultaneously escalating an alert to the human keeper.

By replacing manual roll-calls and chaotic group-chat triage, this setup gives a CHW back an estimated 6 to 8 hours per week while ensuring at-risk members are noticed early.

What I learnt

This project taught me that large communities fail not from a lack of care, but from a lack of structure. When a group scales, the psychological safety of the members drops while the operational burden on the leader skyrockets.

By automating peer matching and streak tracking, Nudger Bee proves that you can make a massive 80-person workspace feel like an intimate, supportive circle. It’s a lesson that applies far beyond healthcare - whether it's an EdTech cohort where students get lost in the chat or a massive B2B community, scaling connection requires intelligent, automated curation.

Built With

• biome
• bolt
• cron
• git
• llm
• mcp
• node.js
• render
• sqlite
• typescript