• 

  Voltwise_UI_MainView

VoltWise — Beyond ON

Your smart home says it's ON. VoltWise checks if it actually happened.

Inspiration

Smart homes are good at accepting commands, but an ON acknowledgment doesn't necessarily mean an appliance is functioning correctly.

Imagine telling Alexa to turn on a heater. The device reports ON, but it draws only 7 watts instead of the expected 1,500 watts. The command appears successful, yet the intended result never happened.

This inspired me to build VoltWise: a smart-home verification system that checks actual electrical behavior instead of blindly trusting device acknowledgments.

What it does

VoltWise is an Alexa+-inspired, MCP-powered smart-home verification prototype built around an Observe → Reason → Act → Verify workflow.

I designed it to compare expected appliance behavior with observed electrical telemetry, diagnose failures, make policy-controlled decisions, and verify outcomes.

My demonstration showcases five capabilities:

1. Silent failure detection: A heater reports ON while consuming only 7 W instead of 1,500 W. VoltWise identifies the mismatch and marks the operation as not verified.
2. Evidence-based diagnostics: VoltWise generates troubleshooting information and a copyable technician handoff.
3. Controlled recovery: When permitted by policy, VoltWise attempts one simulated recovery retry and verifies the resulting 1,500 W measurement.
4. Safety-policy enforcement: When consumption reaches 2,200 W against a 1,900 W ceiling, VoltWise triggers a simulated cutoff and blocks further retries.
5. Energy usage explanation: Through real MCP tool execution, VoltWise explains a household consumption increase from 118 kWh to 142 kWh, distinguishing identified contributors from unattributed usage.

How I built it

I developed VoltWise using Python, FastAPI, React, TypeScript, Vite, and SQLite.

The core implementation uses the official Python Model Context Protocol SDK with MCP protocol version 2025-11-25 over Streamable HTTP.

I implemented a real MCP client/server architecture with nine registered tools supporting verification, diagnostics, recovery, safety decisions, and energy analysis.

I used SQLite to maintain application state and built a React-based Alexa-style conversational interface to demonstrate the complete workflow.

I also added automated backend and frontend tests, documented local setup instructions, and published the source code under the MIT license.

Challenges I faced

My biggest challenge was ensuring that a successful command acknowledgment was never confused with a successfully completed action.

I needed to implement expected-versus-observed power comparisons, maintain workflow state, limit recovery attempts, and enforce safety policies.

Another challenge was making the distinction between real MCP execution and simulated smart-home behavior completely transparent.

I also worked to demonstrate multiple connected workflows clearly within the three-minute video limit.

What I learned

Building VoltWise taught me that meaningful agentic systems need more than conversational interfaces and tool calls.

They need evidence, persistent state, decision policies, controlled actions, and post-action verification.

I learned how to structure MCP tools around real workflows and how to design a system that explains not only what it attempted, but whether the intended outcome was actually achieved.

What's next

I plan to explore integrations with compatible real-world energy-monitoring devices and supported smart-home platforms.

Future improvements could include appliance-specific electrical profiles, expanded diagnostics, and more sophisticated energy attribution.

Before any real electrical control deployment, I would need to address hardware integration, authorization, safety validation, and production-grade failure handling.

Demo and source code

Demo: https://youtu.be/kZ1b0-_O1rU

GitHub: https://github.com/pre170/VoltWise

Technical disclosure: The Alexa-style interface, appliance telemetry, and device actions are simulated. MCP communication, diagnostic logic, and tool execution are real. VoltWise is not a native Alexa+ deployment or a certified electrical safety system.

Built With

• docker
• fastapi
• mcp-python-sdk
• model-context-protocol
• python
• react
• sqlite
• streamable-http
• typescript
• vite