Claude Code Sessions Now Message Each Other

You know that moment when you have two Claude Code sessions open — one working on the backend, one on the frontend — and something breaks in one because of a change in the other? Normally you'd stop, copy the information, switch windows, paste, explain context again. As of v2.1.224, Claude can handle that handoff itself. One session tells the other what changed. No copy-paste required.

I'm Evgeny, and I'll be honest: I wasn't sure I'd use this feature until I realized I spend embarrassingly large chunks of time acting as a messenger between my own sessions. Now I just ask Claude to send the message.

1What is cross-session messaging in Claude Code?

Cross-session messaging lets Claude deliver a text note from one of your running sessions to another. The message is something Claude writes — not your conversation history, not files, just a short piece of text for the other Claude to read. On the receiving end, Claude reads the note and reacts: if it's relevant to what that session is doing, it adjusts; if not, it keeps going. Think of it as Claude sending a sticky note to its counterpart in your other terminal.

2When is this actually useful?

Three situations come up constantly for me. First, a rename or schema change in one session breaks assumptions in another — instead of me noticing and re-explaining, Claude in session A tells Claude in session B before it goes further. Second, I run a long migration or test in session B and ask session A to report back when it's done, so I don't have to check. Third, I work the same codebase from two terminals and ask one session to summarize what it landed so the other can pick up without re-reading the whole diff. You don't have to orchestrate any of this in advance — just ask Claude mid-task.

3How do you ask one session to message another?

You don't need to set anything up. In the session that has the information, just tell Claude what to pass along and who to tell. Claude finds the other session using the /list-agents command (which you can also run yourself to see what's available), picks the right target by name, and writes the actual message. You describe the intent; Claude writes the note.

Tell the session working on the payments API that users.name is now users.display_name

Does Claude decide on its own, or do you have to ask every time?

Both. Claude can send a message without you asking when it notices a change in one session that affects what another session is building — for example, after renaming a function that the other session is calling. It can also reply to incoming messages automatically. And you can always prompt for it: 'tell the other session what we just did' works exactly as you'd expect. You stay in control — Claude won't send your conversation history or files to another session, only text it writes.

4What does an incoming message look like?

When a message arrives, your session shows a 'Message from' row in the conversation. Press Ctrl+O to expand it and read the full text. If your session is idle, Claude Code starts a new turn with the message — same as if you typed something. If you're mid-task, the message lands between tool calls without interrupting the current one. Once delivered, the message counts as a prompt toward your usage, and the receiving Claude can reply the same way.

5Can sessions on different machines message each other?

Yes — sessions on another machine or on Claude Code on the web show up in /list-agents when you're connected to Remote Control. Messages to other machines travel through Anthropic servers (not local-socket-only). This requires v2.1.225 or later for starting a conversation; earlier versions can only reply to a message that arrived first. Sessions inside Docker containers can't reach sessions on the host by default, since they don't share a filesystem — but two sessions inside the same container can still message each other.

6Requirements and availability

Cross-session messaging works on macOS and Linux. It requires Claude Code v2.1.224 or later — run claude update to get it. When a session meets the requirements, messaging is on with nothing to enable. Windows is not yet supported. The feature is not available when using Claude Code through Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry.

Kicker: you've been acting as a messenger between your own AI sessions. Now they can do it themselves. Start a second terminal session this week, ask Claude to coordinate with it on a real task, and notice how much less you have to explain twice.