What We're Building

5,023 catalogued attack vectors across 32 categories for stress-testing AI guardrails

Signed verdicts anchored in an RFC-6962 log, verifiable offline via @omegaengine/verify

Self-assessed framework evaluations with signed, verifiable exports

Emergency stop console for agent traffic

LangChain, LangGraph, CrewAI, AutoGen and more — 13 packages on PyPI plus @omegaengine/mcp on npm

Republishing the framework drop-ins so transport failures deny instead of approve

Open the repository and take billing live

Controls are self-assessed today; a third-party audit is planned — we won't call it certified until it is

Air-gapped self-hosting support beyond today's Docker/Kubernetes configs

Recurring scans with exportable signed reports