Claude Code Adds Cross-Session Messaging on macOS

[MacRumors]

Anthropic has updated Claude Code so that separate coding sessions can message each other directly, removing the need to manually copy context between Terminal windows.

The company shipped the feature in Claude Code version 2.1.224, released on Friday for macOS and Linux. Two new tools power it: ListAgents, which finds other active sessions running on the Mac, and SendMessage, which delivers the message. A session that receives one shows it as a labeled card with a link back to the sender.

According to Anthropic's documentation, only text passes between sessions, not conversation history, files, or permissions. One session can warn another when a change breaks something it depends on, or pass along an answer another session is waiting on, and Anthropic says an incoming message can't approve a permission request or change a session's settings on its own.

New in Claude Code: your sessions can now message each other.Instead of having to re-explain yourself in another session, you can now tell Claude to do it. It sends a summary (not your history or files), and the other session picks it up mid-task. pic.twitter.com/PtNsfXeQXP— ClaudeDevs (@ClaudeDevs) August 7, 2026​

The feature sits alongside Claude Code's existing subagents and Agent Teams tools, but connects sessions a user is already running separately on their Mac rather than ones Claude spawns and manages itself. Messages between sessions on the same Mac stay local and never reach Anthropic's servers. Between different machines, messaging works only as a reply through Remote Control, the feature that connects a Mac terminal to the Claude apps for iOS and Android.

Cross session messaging requires macOS or Linux with Claude Code 2.1.224 or later and isn't available on Windows. It was released alongside 31 other changes in the same release, including a new claude self-hosted-runner command for Team and Enterprise plans.

Article Link: Claude Code Adds Cross-Session Messaging on macOS

 

[kevinon]

I've been using Claude on my iMac and Neo and haven't needed this particular functionality. But that's a useful thing to know.

If anyone has any gotchas about Claude Pro on Mac, or amazing ideas, I'm following along. I am using Claude as an outsourced buddy for my possibly ADHD brain.

 

[Astuces iOS]

yippee more useless ai features !

 

[User 6502]

Does it work with the desktop app or just CLI interface?

 

[marco114]

I've never had a problem before of Claude going into another project if I ask it to. However, I do run everything on dangerously bypass permissions.. I live on the edge.

 

[PixelPanics]

Ah cool, also why does the Claude logo look like a bu**h*le

 

[applicious84]

Ah cool, also why does the Claude logo look like a bu**h*le

I dunno what bu**h*les you're looking at, but maybe start dating some other people? JK!!

Or go to a doc! You've clearly got bigger problems than ai.

 

[mikebenton]

Claude takes about 5x longer than almost any other AI. While it does well when working with very large documents, it does not do well at most other things. I'm just referring to the chat prompt, not Code or any kind of agents.

I also found that many times it takes it upon itself to do a ton of tasks that you never asked for...resulting in token usage. A couple days back I was trying to edit a 60-page bylaws document. Just by asking a question, it would redraft the entire document. After about 3 questions, it told me I was at my limit and had to wait 2 hours before I could continue.

 

[MacMan988]

Thanks but we are fine with Siri

 

[kushner]

Ah cool, also why does the Claude logo look like a bu**h*le

Same reason amazon’s looks like a penis.

 

[erthquake]

Claude takes about 5x longer than almost any other AI. While it does well when working with very large documents, it does not do well at most other things. I'm just referring to the chat prompt, not Code or any kind of agents.

I also found that many times it takes it upon itself to do a ton of tasks that you never asked for...resulting in token usage. A couple days back I was trying to edit a 60-page bylaws document. Just by asking a question, it would redraft the entire document. After about 3 questions, it told me I was at my limit and had to wait 2 hours before I could continue.

Sounds like you may be using Fable. Use Sonnet or Haiku and write your prompt so that you articulate exactly what you want. If anything's vague, the LLMs will make their own decision. Sometimes, it's just better to ask AI to find the section you want to edit. Then copy and paste that text yourself and specifiy the edits you want. And it's probably better to just manually important documents altogether.

 

[ProbablyDylan]

So this does not allow Claude to send messages to subagents after spawning them?

Right now, if a subagent has a question, Claude can't just answer it. It has to respawn the subagent with the new information in its prompt. Kinda annoying.

 

[JungleNYC]

Ever since the Siri announcement last month ChatGPT and Claude have been BEGGING to access my system in various ways.

I use Chat/Claude/Grok combo for work, every day, all the time. But for on-board local file manipulation I am very uncomfortable with these requests. I don't want them reaching into my system. Which is why I like the idea of where Apple is going with Siri AI.

 

[JMalone]

Ah cool, also why does the Claude logo look like a bu**h*le

That’s the Singularity. It is also the Big Bang.

Now you know what the universe is. A giant bu**h*le that began as expansion of energy so that matter could exist so that we could exist so that we could create a gigantic AI so that the universe could know itself. We created God in our own image not vice versa.

 

[timd.mackey]

Messages between sessions on the same Mac stay local and never reach Anthropic's servers.

This is 100% not true in any meaningful sense.

It’s possible the messaging interface is local, but in order for Claude to read the message it is immediately processed by Anthropic’s servers. That’s just how LLMs work.

 

[LincolnNE]

I also found that many times it takes it upon itself to do a ton of tasks that you never asked for...resulting in token usage. A couple days back I was trying to edit a 60-page bylaws document. Just by asking a question, it would redraft the entire document. After about 3 questions, it told me I was at my limit and had to wait 2 hours before I could continue.

(1) - Say “don’t redraft the doc, just answer the question”
(2) - Consider upgrading to a paid tier for more usage capacity

 

[Steve01234]

this is incredibly helpful to move current work to new sessions.big time saver

 

[Wowfunhappy]

Messages between sessions on the same Mac stay local and never reach Anthropic's servers.

Um, what?

It's not a local model. Of course it reaches Anthropic's servers. Otherwise, how would the AI see it?

I think I know what they're trying to say but it seems pretty misleading and also irrelevant.

 

[Wowfunhappy]

So this does not allow Claude to send messages to subagents after spawning them?

Right now, if a subagent has a question, Claude can't just answer it. It has to respawn the subagent with the new information in its prompt. Kinda annoying.

This isn't true anymore. Don't remember when they changed it, maybe within the past few weeks?

 

[Steve01234]

Um, what?

It's not a local model. Of course it reaches Anthropic's servers. Otherwise, how would the AI see it?

I think I know what they're trying to say but it seems pretty misleading and also irrelevant.

for those who don't AI - the sessions write local files. they write notes for AI instance in each session to have quick get up to speed notes on what you are doing.

 

[mashdots]

Ah cool, also why does the Claude logo look like a bu**h*le

thanks, now i can't unsee that

 

[sevoneone]

This isn't true anymore. Don't remember when they changed it, maybe within the past few weeks?

I've been doing it for at least a month. I can ask questions or ask my main/orchestraing session to get an update/pass info to a sub-agent and they'll pass notes.

 

[castlema]

What Maxwell Zeff said in a Wired article remains true; outside of the tech bubble people just do not see any uses for agentic AI, certainly not anything worth paying for. I don't use it and I'm getting along just fine.

Why Normal People Aren’t Using AI Agents

The tech industry is realizing it needs to build agents based on what regular consumers want, not just what its AI models can do.

www.wired.com

 

[JMalone]

What Maxwell Zeff said in a Wired article remains true; outside of the tech bubble people just do not see any uses for agentic AI, certainly not anything worth paying for. I don't use it and I'm getting along just fine.

Why Normal People Aren’t Using AI Agents

The tech industry is realizing it needs to build agents based on what regular consumers want, not just what its AI models can do.

www.wired.com

There are over 2 billion users already and actually there are many users who don’t even know they are using agentic AI and language models because they run in the background as part of the services they already use.

The services you already use from email to streaming already have agents running in the background. Engineers at every major company have already deployed them in almost every popular service for bug fixing, upgrades and business analysis.

The technologies will keep growing until there is 100% coverage. You and “Maxwell Zeff” aren’t going to stop that.

FYI, that does not mean the doom monger stories about mass joblessness are true. Those stories are fake. AI will be everywhere and most jobs will keep existing and there will be many new job types too.

 

[Wowfunhappy]

The services you already use from email to streaming already have agents running in the background.

But are the agents doing things that people actually want? I really don't appreciate having AI in my gmail thank you very much.

Engineers at every major company have already deployed them in almost every popular service for bug fixing, upgrades and business analysis.

Oh, is that what you meant by "running in the background"? While it's true that e.g. Netflix engineers are likely using AI agents to develop the Netflix client, that's not quite the same thing—when I click around the Netflix UI I'm not interacting with an AI agent at all.