Adding Energy Tracking to Open-Source AI Tools with Neuralwatt

How to integrate Neuralwatt with your existing tools

Tools like OpenClaw are quietly driving massive increases in datacenter energy usage.  

‍

Point it at Neuralwatt and you get a full-time AI assistant summarizing emails, drafting PRs, answering Discord . . .

. . . and now you can see exactly how much energy that costs.

‍

So we’re open-sourcing how we use this internally:

👉 https://github.com/neuralwatt/neuralwatt-tools

We're sharing configs, plugins, and scripts to make your stack energy-aware.

‍

🔧 Works with the tools you already use:

These integrations can easily use our latest open-weight models:

  • Kimi K2.5
  • Qwen 3.5
  • MiniMax M2.5
  • …and more

All via OpenAI-compatible endpoints, meaning most tools that support custom base URLs work out of the box.

‍

⚡ Quick start

Grab an API key: https://portal.neuralwatt.com

llm chat -m neuralwatt-kimi-k2.5 -o show_energy true
> What's a hashmap?

A hashmap is a data structure that implements an associative array...

⚡ 3.17kJ | 1.63kW | 7.77s | 881.74mWh | 69.8 tok/s | TTFT 177ms

‍

If you care about where AI is going, star the repo.

We’re building the standard for energy-aware AI.

https://github.com/neuralwatt/neuralwatt-tools

‍

🔗 Links

‍