Guardana
Guardana is an open-source AI security framework that operates offline to scan model artifacts, probe live endpoints, and grade agent execution traces.
Open productWhat it does
About Guardana
Guardana is an open-source AI security framework operating fully offline under the Apache-2.0 license. It provides deterministic build-time static checks for code configurations and model artifacts without requiring network access or external models. For dynamic environments, Guardana executes runtime checks against live endpoints using configurable evaluators, while converting agent runs from tools like PydanticAI, LlamaIndex, and CrewAI into a single unified trace model for evaluation.
- Availability:
- launched
- Access:
- free
- Format:
- software
- Platform:
- cli
Core capabilities
Static Build-Time Inspection
Runs deterministic static checks on model artifacts and code configurations without requiring external models or network connectivity.
Dynamic Endpoint Probing & Evaluation
Interacts directly with live endpoints to execute runtime security checks, evaluating policies with versioned, swappable evaluators that deliver detailed outcomes and rationales.
Keep discovering
Related debuts
Continue with products that solve a similar problem or serve the same kind of maker.
