Kastra

Meet Kastra

Kastra adds policy checks to actions taken by AI systems. Teams define what an agent may do and place authorization checks before supported prompts, tool calls, shell commands, and API requests reach their targets. Rules can allow an operation, deny it, or require a person to approve the step.

Its Edge component applies controls to supported coding agents on developer computers, while a shared console manages policies and decision records. Recon can inspect existing agent history and suggest policies based on the activity it finds. Deployment options include managed and customer-operated environments. Effective coverage depends on connecting the relevant execution paths and configuring policies for the team's actual tools and workflows.