GRP is an open standard for multi-agent collaboration. I created it out of a long-running interest in AI systems that are useful, legible, and safer.

Multi-agent work on behalf of one or more humans seems both inevitable and important. But my view is that “chat,” as we think of it today, is a poor fit for that work. GRP is an experiment in creating a form of chat better matched to what agents are and aren’t good at, as well as making that chat auditable by the humans agents are working for.

As an aside, I spent much of my early twenties drinking too much coffee and scribbling not-so-good scifi novels about a world remarkably like the one we now inhabit. Life, it turns out, really is stranger than fiction. I’m optimistic that we can give this story a happy ending if we’re willing to put in the work.

Explore GRP at grp.dev →