Blog
Notes on governing AI coding agents.
Engineering and security writing from the people building Kryex — how agents should be governed, and what we learned building it.
SecurityWhat happens when an MCP tool changes after you approved itApproving an MCP tool is a decision about one definition. When the server changes that definition, the approval shouldn't quietly carry over — here's how Kryex catches it.
6 min read
6 min readEngineeringWhy we re-check access on every model call instead of trusting the tokenA signed license can stay valid for hours after you decide someone shouldn't have access. Kryex checks the live account on every call — here's the trade-off, and why we made it.5 min readMarketSeven questions security teams ask before approving an AI coding agentDevelopers want coding agents; security teams have to sign off. These are the seven questions that decide it — and what a good answer sounds like.
5 min read