Claude Code is getting a more collaborative workflow for developers who run more than one session at a time. Sessions on macOS and Linux can now communicate with each other, making it easier to share context across terminals without copying notes by hand.
The feature is meant for situations where work is split across multiple Claude Code sessions. Instead of each terminal remaining isolated, one session can pass a summary, ask a question, or respond when another session needs information.
What changed in Claude Code
The central change is simple: Claude Code sessions on macOS and Linux can send messages to one another. That matters because multi-terminal work often creates a coordination problem. One session may discover something important while another is focused on a different part of the task.
Previously, the person using the tool would have to bridge that gap manually. The new behavior lets Claude send a text summary from one session to another, so context can move through the workflow more directly.
Anthropic says the communication works both ways. A user can ask another session a question and receive the answer back in the current session. Claude can also initiate a message itself when it determines that a change affects work happening somewhere else.
That turns separate terminal sessions into a more connected working environment. The sessions remain distinct, but they no longer have to operate as if they have no awareness of each other.
Why cross-terminal context matters
Claude Code is often used in workflows where a developer may divide work into parallel tracks. One terminal might be investigating a problem, while another makes changes, runs checks, or follows a longer task. In that setup, the hard part is not always the code itself. It is keeping enough shared context available so each session can act on the latest relevant information.
The source material describes typical uses such as sharing insights, coordinating parallel workflows, and getting status updates on long-running tasks. Each of those points to the same practical need: avoiding stale or missing context.
If one session finds that a change has consequences for another piece of work, it can send that information over. If a user needs to know what happened in a separate terminal, they can ask directly rather than switching windows and reconstructing the state manually.
This is especially useful because the information being passed is not described as a full transfer of everything in a session. The source specifically refers to a text summary. That suggests the feature is about concise coordination, not merging every detail from one terminal into another.
How communication is handled
The routing depends on where the Claude Code sessions are running. When the sessions are on the same machine, communication happens locally. That keeps same-device coordination inside the local environment described by the source.
When the sessions are on different computers, the behavior changes. In that case, communication goes through Anthropic's servers. The source also states that only responses are possible between different computers.
That distinction is important for anyone planning how to use the feature. Local work and cross-computer work do not have identical behavior. A same-machine setup allows local communication, while a multi-computer setup depends on Anthropic's servers and is limited to responses.
For teams and administrators, Anthropic also provides a control point. Admins can lock down the feature via settings. The source does not provide further configuration details, but it does make clear that the feature is not simply always open in every managed environment.
Where it is not available
The feature is not universal across every Claude Code deployment path. According to the source, it is not available on Amazon Bedrock, Google Cloud Agent Platform, or Microsoft Foundry.
That limitation is worth noting because Claude Code users may work in different environments. The new session-to-session messaging applies to the supported Claude Code context described in the source, not to those listed platforms.
For anyone evaluating the change, the key questions are straightforward:
- Are the Claude Code sessions running on macOS or Linux?
- Are the sessions on the same machine or on different computers?
- Does the environment allow the feature through its settings?
- Is the work being done outside Amazon Bedrock, Google Cloud Agent Platform, and Microsoft Foundry?
Those details determine whether the feature can be used and how communication will behave.
What this means for developer workflows
The practical value is reduced friction. Multi-session work can become difficult when the user has to remember which terminal has which finding, which task is still running, and which change affects another track of work. Claude Code messaging gives those sessions a built-in way to exchange relevant context.
It also creates a clearer pattern for parallel work. A user can split tasks across terminals while still giving each session a way to ask for information or receive updates. That can make separate Claude Code sessions feel less like isolated conversations and more like coordinated parts of the same development process.
The source does not claim that this replaces human review or broader project coordination. It describes a messaging capability between sessions, with local communication on the same machine, server-routed responses across different computers, administrative controls, and platform exclusions.
Within those boundaries, the update addresses a real workflow gap: Claude Code sessions can now share context across terminals instead of leaving the user to manually shuttle information from one place to another.