Claude Code sessions now gossip across terminals—and that changes workflows
Curated by the Inblix editorial team
Anthropic just gave Claude Code a feature that sounds small but rewires how developers will use the tool: sessions can now send messages to each other. On macOS and Linux, a Claude instance in one terminal can fire off a text summary to another session—no copy-paste required. The communication flows both ways. You can ask a question and get an answer back directly, or Claude can proactively message another session when a code change might affect what that other instance is working on.
The practical implications are immediate. If you’re running parallel Claude sessions on different tasks, they can now coordinate without you playing telephone operator. Sharing insights, checking on long-running jobs, or handing off context between workflows all become programmatic rather than manual. On the same machine, this happens locally. Across different computers, the messages route through Anthropic’s servers, though cross-machine communication is limited to responses only—you can’t initiate from afar.
Anthropic’s documentation frames this as a productivity multiplier for developers juggling multiple Claude instances, and that’s exactly right. What’s notable is that Claude can initiate the conversation unprompted. If Session A makes a breaking change to a shared dependency, it can warn Session B before you even realize there’s a conflict. That level of agent-to-agent awareness inches closer to the multi-agent systems researchers have been prototyping in labs. It’s not full autonomy, but it’s a meaningful step past the single-session silo most coding assistants operate in.
There are guardrails. Admins can disable the feature entirely via settings, and it’s not available on Amazon Bedrock, Google Cloud Agent Platform, or Microsoft Foundry deployments. Anthropic is keeping this tight to their own infrastructure for now. The server-routed cross-machine path also raises questions about latency and reliability that the docs don’t fully address yet—something to watch as teams start stress-testing this in real projects.
💡 Key Takeaways
- Claude Code sessions on macOS and Linux can now send messages to each other, with Claude able to initiate communication when it detects relevant changes
- Cross-machine communication routes through Anthropic's servers and is limited to responses only, while same-machine messaging stays local
- The feature is unavailable on Bedrock, Google Cloud Agent Platform, and Microsoft Foundry, and admins can disable it entirely via settings
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.