All activity
Is your claude code doing funky things? Is your agent calling the wrong tools or following unwanted instructions? Krnel-graph allows agent developers to build reliable and secure agents.
GitHubWhere reliable and safe agents are forged
Peyman Faratinleft a comment
There are something like 150 people in the world who know how to train frontier models. And 0 who know how they work. This makes reliability and safety a game of chance. Our goal is to bring frontier AI toolings that look at model internals to the larger agent developer community, so that developers can detect unwanted behaviors and control them, without having to be an AI researcher.
GitHubWhere reliable and safe agents are forged