Billie — Energy Cost Intelligence Agent

Your energy bill shouldn’t be a mystery. Billie spots costly usage before it adds up, helps you act, and explains where your money went.

Comment

Inspiration

We started with a simple question: when someone's electricity bill suddenly increases, why should they have to figure out the reason themselves? We wanted to build something that could investigate the problem, explain what was happening, and suggest what they could do about it. That's how Billie — Energy Cost Intelligence Agent came to life.

What it does

Billie helps users understand household energy usage through an interactive dashboard and an AI agent. It brings together an energy overview, virtual appliances, simulated bill comparisons, energy alerts, and cost-saving opportunities.

Users can change appliance settings in the virtual smart home and see how energy consumption and estimated costs change. They can also ask Billie questions like, “Why is my electricity bill high?” Billie uses energy-analysis tools to investigate the available data and explain its findings. Its MCP tool trace makes the underlying tool activity visible.

How we built it

We built Billie using Python, a locally hosted Ollama language model, MCP tools, and a web interface. The Python backend connects the AI agent to energy-analysis functions, while the virtual smart-home environment provides simulated appliance states and energy readings.

We also built dashboard sections for energy insights, alerts, recommendations, demo controls, and MCP tool tracing. This lets us demonstrate how conversational AI can work with tools and structured data rather than relying only on generated text.

Challenges we ran into

Getting the different components to work together was one of our biggest challenges. We encountered MCP connection failures during testing and a Windows command-execution helper issue that was resolved by restarting Windows.

We also had to pay attention to consistent energy-cost calculations, differences between energy charges and fixed charges, and whether simulated actions actually changed the application state. These challenges pushed us to inspect tool results, test behavior, and avoid assuming an operation had succeeded without verification.

Accomplishments that we're proud of

We're proud of bringing conversational AI, MCP-based tool interaction, energy analysis, and a virtual smart home into one project.

Billie's goal goes beyond answering questions. It is designed to investigate energy usage, identify unusual consumption, explain estimated cost impacts, recommend potential savings, and verify simulated interventions. We're also proud that the virtual environment lets us demonstrate these ideas without requiring physical smart-home hardware.

What we learned

We learned that an AI agent is much more than a language model connected to a chat interface. It needs relevant tools, reliable communication, consistent data, meaningful evidence, and safeguards around actions.

We also learned the importance of transparent estimates and clearly distinguishing simulated energy readings from real household measurements. Debugging the integration challenges gave us a better understanding of how to build and test tool-using AI systems.

What's next for Billie — Energy Cost Intelligence Agent

Our vision for Billie is to make household energy costs easier to understand and energy-saving decisions easier to explore. We want the experience to remain transparent, with clear explanations of cost estimates, visible tool activity, and verified outcomes for simulated actions.

Billie demonstrates how an AI agent can connect a user's question to energy data, investigation, and practical recommendations in one place.

Built With

  • ai-agents
  • css
  • data
  • energy-management
  • html
  • javascript
  • large-language-models
  • mcp-(model-context-protocol)
  • ollama
  • python
  • rest-api
  • smart-home
Share this project:

Updates

Submission history