Our Claude Code sessions can now talk to each other, and that changes everything

Anthropic has just taken a very interesting step inside Claude Code. As of version 2.1.224, the sessions on macOS or Linux can send messages to each other. A somewhat peculiar behavior, yes, but one that completely changes how we can use the tool, because the shared context is crucial for working efficiently in parallel and makes Claude an even more autonomous tool.

Visual Studio Code Download

Several sessions working like a real team

As Anthropic explains in the official documentation, one session can locate the others and send them information, decisions, or questions. So, if we have one session working on an API, for example, and another reviewing the interface, both can coordinate changes while continuing to move forward.

The idea does in fact recall the Claude Tag system, where Claude participates proactively in workflows thanks to the information it receives from a conversation. Here the coordination happens between sessions, automatically and with the power of Claude Sonnet 5 and the rest of the models, but the principle is similar.

More autonomy, but keeping our permissions

The system is designed to exchange summarized text messages, but each session always keeps its permissions. Any sensitive action still requires our approval and does so independently of the conversation the contextual message came from.

With features like Claude Security or on larger projects, task division is essential, and sharing context between them completely changes how we use Claude Code.