This is a submission for the OpenClaw Challenge.
What I Built
I'm from Rajasthan, India — where villages are far apart and the nearest doctor can be hours away. Rural and tribal communities here use WhatsApp daily, but have no easy access to basic health guidance.
So I built Aarogya Saathi (आरोग्य साथी) — a WhatsApp-based AI health assistant powered by OpenClaw, deployed on AWS EC2, designed specifically for rural India. It speaks Hindi, gives first aid guidance, handles emergencies, and works 24/7.
The problem it solves:
- Nearest doctor is 20-50 km away in rural Rajasthan
- People only understand Hindi, not English apps
- WhatsApp is the only technology they use daily
- No awareness of emergency numbers like 108, 104
Aarogya Saathi bridges that gap — AI-powered, always on, completely in Hindi.
How I Used OpenClaw
I deployed OpenClaw on an AWS EC2 Ubuntu 22.04 instance and connected it to WhatsApp using the built-in QR pairing channel. Mistral AI powers the language model.
Infrastructure:
- AWS EC2 Ubuntu 22.04 (24/7 always on)
- OpenClaw as the AI gateway
- WhatsApp channel via QR pairing
- Mistral AI as the language model
The most important part was the custom agent prompt — tuned to behave like an ASHA worker (Accredited Social Health Activist):
You are Aarogya Saathi (आरोग्य साथी), a trusted health assistant
for rural and tribal communities in Rajasthan, India. Always reply
in simple Hindi or Hinglish. NEVER replace a doctor. For emergencies
ALWAYS mention 108 (Ambulance) and 104 (Health Helpline). Give
practical first aid for fever, dehydration, snake bite, diarrhea.
Keep answers short and warm like an ASHA community health worker.
What OpenClaw handled automatically:
- WhatsApp QR pairing in under 1 minutes
- Agent workspace auto-bootstrap on first message
- Systemd daemon — survives server restarts 24/7
- Per-user session memory built-in
- EC2 deployment with zero extra configuration
Demo
Test conversations on WhatsApp:
- General Introduction to Health Assistant
- 🤒 "Bukhaar hai 102 degree, kya karun?" → Hindi first aid steps
- 🐍 "Saanp ne kaata emergency kya karun?" →
- 🌿 "Aaj ka health tip do" → Daily health tip in Hindi
- ☁️ EC2 instance running 24/7 on AWS
What I Learned
Biggest surprise: OpenClaw's agent auto-bootstraps its own workspace on the very first message — it created identity files, health log directories, and memory files completely on its own. That was impressive.
Key challenges:
Hindi system prompt tuning took multiple iterations — the warmth and simplicity of an ASHA worker is hard to capture
WhatsApp QR pairing on a server (no display) needed careful terminal handling
Mistral AI free tier has rate limits — had to be mindful during testing
Key takeaway: The hardest part wasn't the tech — it was writing a system prompt that feels human, warm, and trustworthy to someone in a rural village who has never used AI before.
This project showed me that AI accessibility isn't just about language — it's about tone, simplicity, and meeting people where they already are (WhatsApp).
ClawCon Michigan
I did not attend ClawCon Michigan — I'm based in Rajasthan, India! But building this project made me feel connected to the OpenClaw community from across the world. 🇮🇳
Top comments (2)
Great work. I believe, it's worth to mention about the app features.
Features
Also mention about the problems and solutions.
Problem:
Solution:
Thanks for the incredibly insightful feedback, Ranjan! You've pointed out the exact 'make-or-break' factors for an AI in the medical domain.