
I built a local AI workspace for Windows and finally released it
After about a year of building, testing and constantly finding things to improve, Nyx AI is finally out. It is available in 240 markets and 6 languages.
Nyx AI is a local first AI workspace for Windows for chat, coding, documents, previews and automation.
You can run models locally through Ollama or connect OpenAI, Anthropic, Ollama Cloud or your own compatible endpoint. No Nyx AI account is required for local use.
One area I spent a lot of time on was agent safety.
In Nyx AI Lab the model can work with files, run shell commands and Python, use Git, run builds and tests, and preview PDF, DOCX, XLSX and PPTX files.
Locked execution uses Windows AppContainer isolation where supported, alongside workspace file controls, command risk checks, approval controls and native Windows confirmations for certain higher risk actions.
There is also a built in privacy sanitiser designed to detect and redact recognised sensitive information before messages are sent to non local models. It is best effort and cannot guarantee that every sensitive value will be detected.
No sandbox is completely secure and these controls reduce risk rather than eliminate it. The limitations and residual risks are documented in the public threat model, and the legal and technical documents are available inside the app.
Would genuinely appreciate anyone trying it and telling me what you think.
Website
[https://nyxai.uk
Microsoft Store
)Microsoft store nyx ai