Chrysent Bastion

Chrysent Bastion protects the execution boundary around AI agents. When an assistant has access to tools, local machines, remote systems, or MCP servers, Bastion inspects attempted tool calls and can reject unsafe actions before they touch sensitive resources.

It comes with sane defaults for common risks, including dangerous commands and protected paths, plus configurable policies for organization-specific rules. Bastion can explain a rejection when transparency is useful, conceal the reason when revealing policy would help an attacker, and keep an internal audit trail for Chrysent administrators.

For teams with advanced requirements, Bastion includes a policy engine and tester for developing custom controls before they run in production. Organizations can define and validate their own rules through Chrysent’s policy DSL, then run them through Gateway as part of the broader AI control plane.