Inspiration
Access to clean water is still a major challenge across many African communities, where most people only have basic phones and limited or unreliable internet. We wanted to prove that advanced AI can deliver real impact in these environments — not by replacing what exists, but by augmenting it with an intelligent, autonomous layer. The inspiration behind FlowAgent is simple: use AI + SMS + IoT to make clean water available anywhere, even where technology is scarce.
What it does
FlowAgent allows anyone to buy clean water using a simple SMS, no apps or internet needed.
When the user pays, FlowAgent uses a Nullshot AI Agent to verify the transaction on-chain and securely trigger a real IoT water pump via an ESP32 T-Call. Everything is logged, monitored, and visualized in a Nullshot-powered web dashboard.
FlowAgent transforms any water kiosk into:
an autonomous AI-operated utility,
transparent and tamper-proof,
cashless,
reliable even with 2G coverage, and
accessible to all, including communities without smartphones.
How we built it
We built FlowAgent by combining Nullshot, Web3, and IoT hardware:
Nullshot AI Agent
Responsible for reasoning, validation, and decision-making
Triggers hardware actions via MCP tools
Interfaces with blockchain for payment confirmation
Custom MCP Tool
Communicates directly with the ESP32 (LilyGO T-Call SIM800L)
Sends ultra-light (1–5 byte) commands
Handles pump activation and status reporting
Python Listener
Receives SMS payment confirmations
Forwards data to the Nullshot Agent
Logs events and handles fallback logic
ESP32 Water Controller
Activates pump based on commands
Tracks volume, timing, and errors
Sends back telemetry
Nullshot Web App
Dashboard for operators
Real-time logs, pump states, payment records
Manual override + sensor visibility
Blockchain Layer (Base L2 / or any EVM)
Payment verification
Tamper-proof transaction logs
Everything works together to build an AI-native autonomous water system.
Challenges we ran into
Designing a reliable pipeline with zero internet for the end user
Making SMS payments compatible with real-time blockchain verification
Building an MCP tool that can communicate with hardware in extremely lightweight packets
Ensuring pump activation is safe, repeatable, and tamper-resistant
Syncing hardware state with Nullshot’s web dashboard
Optimizing the AI agent to make deterministic decisions in a low-connectivity context
Accomplishments that we're proud of
Built a fully autonomous water-control system powered by Nullshot AI
Demonstrated real hardware control using MCP tools
Created a fully functional low-connectivity payment → AI → IoT → pump pipeline
Achieved end-to-end verification from a simple SMS all the way to a real water pump
Developed a clean, instantly deployable web dashboard powered by Nullshot
Proved that AI can operate in environments with 2G-only connectivity
What we learned
How to integrate traditional telecom (SMS) with modern Web3 systems
How powerful Nullshot is for creating AI-native autonomous agents
How to architect reliable IoT systems under extreme constraints
The importance of deterministic AI reasoning in real-world infrastructures
That simplicity (SMS + AI + IoT) can actually produce the most robust solutions
What's next for FlowAgent
Deploy pilot units in Burundi and East Africa
Add prepaid NFC card support for offline villages
Integrate solar-powered sensors for water-level monitoring
Expand the dashboard with predictive analytics for maintenance
Add community-level DAO governance for pricing and transparency
Partner with NGOs, water operators, and humanitarian organizations
Scale FlowAgent to support agriculture, irrigation, and remote utility management
Log in or sign up for Devpost to join the conversation.