ChameleonOS

A personal AI operating system for Windows. Runs on your machine.

Get it from Microsoft Store Learn more · main site

$49.99 once. No subscription. No API keys.

ChameleonOS is the environment. It holds the memory, the persistence, and the working state that survive across sessions, so the things you do with an AI are not lost when the chat ends. It is not the assistant inside the chat. It is the layer beneath it.

Two systems, one reason

ChameleonOS can talk to AI models two ways. They look similar from the outside. They are not, and the difference is about compliance.

Universal Automate

Some web models do not allow programmatic automation. For those, ChameleonOS uses Universal Automate. That process does not automate the model. It automates your own actions — moving the mouse, typing, clicking, reading the reply — exactly the actions a person would take. Because it reproduces your actions rather than calling the model programmatically, it does not violate the model's automation policy.

Residents

For models that do allow automation, or that you are entitled to automate, ChameleonOS runs a resident in the background. It works without bringing windows to the front — quieter and faster, because it does not have to reproduce every human action.

That is why both exist. Neither replaces the other. Universal Automate stays compliant and reaches any model you can use in a browser. Residents are for the models that permit automation. Same product, two paths.

What runs on it

Free Agent — the agent layer

Free Agent is the agent that runs on ChameleonOS. It is the service layer that connects everything: the memory system, the automation system, the ability to drive tools and act on your machine. It is called a free agent in the sports sense — not tied to any one team. It is not locked to Gemini, or ChatGPT, or DeepSeek, or Ollama. You pick the team. It plays. It can work with more than one model at once.

Where it came from

ChameleonOS was built over seven months by one user, without a code background, working with AI on commission. It was not designed in a single plan. It grew the way a tool grows when someone actually needs it: hit a problem, try something, keep what works.

Store information