kontext

Kontext is a runtime governance tool designed for AI agents that enables local policy decisions, pre-action enforcement, and maintains an authorization ledger. It captures tool-use events, evaluates policies prior to action execution, and provides mechanisms for recording decisions, facilitating a balance between security and agent utility. Key features include local enforcement of policies, an observation mode for testing policy impacts, and compatibility with various execution environments, ensuring that security actions are seamlessly integrated into AI workflows without constant reliance on external services.