This project is an experimental research prototype. It is not a certified production security platform and should not be deployed as one without independent validation, threat modeling, and environment-specific testing.
NomaanOS-Core provides a local-first execution and orchestration layer for research experiments focused on:
- offline AI execution
- host telemetry and anomaly observation
- cryptographic audit trails
- disconnected or low-trust coordination patterns
- Python 3.8+
- Linux, macOS, or Android/Termux
- Git
git clone https://github.com/NomaanOS-Dev/NomaanOS-Core.git
cd NomaanOS-Core
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install -r requirements-dev.txtpython nomaanos.py --status
python nomaanos.py telemetry
python nomaanos.py chain-verify
pytest -qLinux / POSIX host
↓
ShieldSOC telemetry layer
↓
NomaanOS-Core orchestration/runtime
↓
EvidenceLedger audit chain
↓
GhostNode / disconnected coordination layer
nomaanos.py— CLI and operational entry pointapi_server.py— local HTTP service for status and telemetry endpointstui_dashboard.py— terminal dashboard for interactive viewingsrc/engine.py— fail-closed execution layersrc/orchestrator.py— orchestration logic for request processingtests/— validation and behavior checks
- no external security audit or formal compliance review
- not a validated enterprise deployment platform
- cryptographic assumptions are local and experimental
- multi-node role testing is still limited
- some modules are research scaffolding rather than hardened production services
pytest -qpython -m py_compile nomaanos.py api_server.py tui_dashboard.py
ruff check .See SECURITY.md for responsible disclosure expectations and threat-model notes.
See CONTRIBUTING.md for development expectations and review workflow.
MIT License — see LICENSE.
- Nomaan Khan
- GitHub: https://github.com/NomaanOS-Dev
This repository is distributed as an experimental research project. It is provided as-is for local experimentation and engineering exploration. It should not be treated as a deployable production security platform without independent validation.