Claude Code Now Shares Context Across Terminals

Anthropic quietly rewired how Claude Code handles multiple terminal sessions running at once. Until now, anyone juggling several instances of the coding assistant had to manually copy context from one window to another, pasting error messages or status updates by hand whenever one task depended on another. That workaround disappears with the new update, which lets Claude Code sessions on macOS and Linux message each other directly.

The feature targets a specific pain point for developers running parallel workflows, say, one session refactoring a database schema while another builds the API that depends on it. Rather than switching windows and relaying updates yourself, one Claude session can now flag changes to another the moment they happen, or answer a direct question fired over from a separate terminal.

Anthropic built in some structure around who gets to talk to whom and how. Location matters: sessions on the same machine communicate differently than sessions split across separate computers. The company has also given system administrators a way to control access to the feature entirely, and it's not rolling out everywhere at once. Bedrock, Google Cloud, and Microsoft's Foundry platform are sitting this one out for now.