AI Assistant for NIST 800-171 Assessments
RAG AI Specialist is an experimental Python framework I started in April 2025 to explore how retrieval-augmented generation could support NIST SP 800-171 cybersecurity assessments. The idea was an AI assistant that works from structured control data and evidence an assessor controls, instead of guessing from general training.
What it explores
- Structured compliance data. A
NISTControlsclass loads control definitions from YAML and looks them up by family and identifier, keeping the reference data separate from the application logic. - Plugins for evidence collection. A plugin manager discovers classes built on a shared plugin base, so each kind of system (workstation, cloud environment, network appliance) can have its own collector.
- Cooperating agents. Network and peer-management modules, with MQTT dependencies, explore letting specialised components talk to each other.
- Careful handling of sensitive data. Design goals include sanitised retrieval data, tamper-evident audit history and controlled sharing between agents. These are goals in the proposal, not finished features.
Where it stands
The repository is a set of early-stage code, configuration experiments and architecture notes. Some parts are working foundations; others are placeholders or proposals. It is not a finished or certified NIST assessment platform, and it does not replace a qualified assessor.
It also became the home for related work on local-first and field-deployable AI. The repo now holds documentation, hardware and scenario examples for Solar Sentinel, alongside Docker files, dashboard code and notes on local AI and RAG integration.
Read the build story
RAG AI Specialist: Building an AI-Powered NIST 800-171 Audit Companion is the full retrospective, including what I would change if I rebuilt it today.
Want something like this for your business? See full-stack and custom development, or book a call.