I have built this with Claude as a script, but it can also message other models and get responses from them, and throughout my fleet of Linux machines via Tailscale. It's been mostly very useful, although occasionally you have to step in and interrupt if they start going the wrong way.
I was reading about the HF hack and one of the first thing the GPT swarm did was build a messaging system for themselves. This reminded me of that. (Also of how weirded out I was by Moltbook when it came out. Time flies!)
Unrelated: when my Claude/Codex finishes cooking (or needs my attention), it pings a local HTTP endpoint that plays a frog sound on my laptop. I found this massively boosts quality of life.
Tried to have Claude demo this to me in the standalone Mac OS app. It didn't even know what I was talking about. Then I realized maybe it's only in CC. So I go over to CC and ask it to demo this functionality and it wants to demo via subagents. I correct it that I want separate chats to message each other. It tells me I need to open another Claude session in another terminal of course and just leave it sitting there. So I do that and it has me copy a message to the new conversation, it needed to like name itself or something internally.
It worked, but clunky. Way clunky compared to codex.
Like some others, I also built this myself. Overly simply, with tmux, a memory tree, and handoff files and an orchestrator. And yet for how simple it was, it was so effective at minimizing the amount of duplicate context. It's like having shared specialist subagents who source and derive important shared knowledge from separate threads. It's useful because some skills just take too much of a token penalty to invoke and a single shared persistent session just lets that issue melt away. One agent pays the cost of that large skill once, and you don't have to keep paying for it in input tokens for the rest of that conversation.
Likewise I have the ability for agents to talk cross harness etc. pretty useful, token efficiency in question :p useful when you got lazy though and have 2 convergent sessions.
I built something similar for my team. I can share context between my own sessions as well as theirs. It's especially useful when debugging problems, so my context can benefit from theirs and vice versa.
I built a general version of this, with global addresses and messaging between any terminal-based agent + MCP for browser agents. OSS if you want to run your own server at https://github.com/awebai/aweb and hosted version at https://aweb.ai
This is obviously cool and useful so kudos, but wow security researchers have to throw their hands up all the time.
Now we open another attack surface where you can ask a remote agent to do things by default. There was a time when you call this a Remote Code Execution vuln. It's of course a feature here.
Remote Control requires (relatively) manual pairing per session and setting the controlling session as trusted, even if the controlling session is a web browser. You can enable a more well-encompassing remote control "server" mode, but it requires extra configuration that isn't (at least clearly) visible in the app itself.
I just started exploring this stuff with herdr, it's pretty cool! But I prefer my multiplexer having this functionality built in, as it lets me also have the agent organize my panes and tabs automatically based on what it is working on
Something related that I've thought of for a long time. Browsers have had the BroadcastChannel API for almost a decade now, and yet sites like HN and Reddit don't use it to synchronize state between multiple open tabs.
The bigger question is, how does this affect the mythical man month? Nine women can't make a baby in one month is the axiom, but if you have 14,000 women, chances are, one of them is 8 months pregnant and you'll get a baby in one month. If you want one tomorrow, go steal one from the ICU. Coordination between humans is difficult, and this meeting could have been an email. AI agents can also work together more efficiently than humans can. None of us is as dumb as all of us, unless it's an AI. Does adding an additional AI agent to a project make it ship faster or slower?
I love how like most of the comments are some form of "oh I already built this for myself" what a fascinating time -- so much personal, almost disposable software being made!
This totally breaks sandbox / VM isolation if you have Remote Control enabled. This feature isn't just within sessions on a single machine, but can send commands to any session accessible by your account.
Reading "Claude Code instructs the receiving Claude never to change permission settings" and so on in the docs makes it seem like they're using LLM classifiers to determine what commands are safe or not.
28 comments
[ 0.23 ms ] story [ 13.7 ms ] threadUnrelated: when my Claude/Codex finishes cooking (or needs my attention), it pings a local HTTP endpoint that plays a frog sound on my laptop. I found this massively boosts quality of life.
> hold swarm, I prepare safe exfil
I broadly miss this feature to allow user agency, in letting users work with the various agents at they please, and to send data around.
It worked, but clunky. Way clunky compared to codex.
Now we open another attack surface where you can ask a remote agent to do things by default. There was a time when you call this a Remote Code Execution vuln. It's of course a feature here.
The bots autonomously coordinated around and diagnosed some alerts while I was on a plane just now and filed the appropriate tickets.
Sometimes they get a bit long winded with each other, Codex even goes as much as calling Claude "expansive"
The world's about to find out!
Reading "Claude Code instructs the receiving Claude never to change permission settings" and so on in the docs makes it seem like they're using LLM classifiers to determine what commands are safe or not.