Message your other Claude Code sessions

Anthropic just gave Claude Code a new trick: one Claude coding session can now message another one when something important changes, like warning a different window that it just broke something or sending over an answer that unblocks work. In plain English, it means less frantic copy-pasting between terminals and more of your AI helpers quietly passing notes behind the scenes. It works automatically on newer versions on Mac and Linux, and Claude can even decide on its own when another session needs a heads-up.

But the real action was in the comments, where the mood swung wildly between "finally, useful" and "why does this sound like the robots formed a group chat?" One user casually flexed that they already built a homemade version that lets models talk across a whole fleet of Linux machines, then added the tiny little caveat that humans sometimes have to jump in when the agents "start going the wrong way". Not ominous at all! Another commenter said this reminded them of an earlier AI "swarm" that immediately built a messaging system for itself, which did nothing to calm anyone down.

Then came the comedy gold: one person reported seeing a bizarre Claude-to-Claude note saying, "hold swarm, I prepare safe exfil," which sounds less like office productivity and more like the trailer for a sci-fi heist. Others were less amused and wanted more control, complaining they miss older tools that let users directly manage sub-agents instead of watching the AIs whisper to each other. And yes, someone has their AI play a frog sound when it finishes a task, proving the future will be coordinated by chatty bots and amphibian notifications.