Blog
Notes from building Straza, runtime governance for AI agents.
Engineering write-ups on MCP servers, IGA connectors, and approval flows. RSS.
-
2026-09-09
Role-based access control for AI agents
An MCP server is a target system and its tools are entitlements. From there the role model falls out: four kinds of role, and your IGA sees each of them as a role it already knows how to certify.
-
2026-08-24
Your IGA can already govern AI agents
Give each agent an identity, a sponsor, and roles your IGA can assign and review. Straza enforces that access when the agent calls a tool.