# Claude Code sessions can now message each other By Steven Van · 2026-08-07 Claude writes a summary itself so a second terminal picks up mid-task, on v2.1.224+ for macOS and Linux. Anyone who runs multiple [Claude]() Code sessions knows the ritual: something important happens in terminal one, so you switch to terminal two and re-explain the whole situation from scratch. Anthropic just removed that step. Claude Code sessions can now message each other directly. You tell Claude to fill in the other session, it writes a summary itself, and the receiving session picks the message up mid-task without you touching a second keyboard. [Source post](https://x.com/i/status/2085817074816070014) ## Claude writes the message, not you The mechanics are refreshingly hands-off. Claude uses two tools under the hood: `ListAgents` to discover which of your sessions it can reach, and `SendMessage` to deliver text to one of them by name. You never call either tool yourself. A prompt as loose as "explain what we just did to the session working on the payments API" is enough, and Claude composes the actual message. It can even send one unprompted, for example after making a change it knows affects what another session is building on. Delivery is polite by design. If the receiving session is mid-task, the message is read between tool calls so a running operation is never interrupted. If the session is idle, the message starts a new turn. You can see every reachable session with the `/list-agents` command, and sessions answer to names you set with `/rename` or that Claude Code derives from the working directory. ## A summary crosses the wire, not your history The privacy boundary is the detail worth knowing. What travels between sessions is a piece of plain text one Claude writes to another, never your conversation history and never your files. On the same machine, messages move over a per-session socket and do not pass through Anthropic's servers at all. Anthropic's docs give a flavor of what actually arrives: "Schema migration finished: the new column is tenant\_id, and rebasing on main is safe now." Short, specific, and enough for the other session to keep moving. Incoming messages are also deliberately powerless. A message from another session cannot approve a pending permission prompt, cannot change settings or configuration, and any slash command inside it arrives as inert text. If acting on a message requires a permission the receiving session does not have, you see the same prompt you always would. Claude Code even throttles message loops between sessions so two agents cannot spam each other into a runaway exchange. ## Built for parallel worktrees and overnight jobs The obvious beneficiaries are people running Claude Code in parallel. When several sessions work the same repository in separate worktrees, one session can announce what just landed so the others stop building on stale assumptions. A long migration or test run can report back to the session you are actually watching. And when one session settles a question another is blocked on, the answer moves across without you playing messenger. Messages can even reach your sessions on other machines or on Claude Code on the web through Remote Control, though across machines Claude can only reply to a message, not start the exchange. ## Where it works and how to rein it in Cross-session messaging requires Claude Code v2.1.224 or later and runs on macOS and Linux, including WSL 2, but not native Windows. When a session meets the requirements it is on by default with nothing to enable. Control is granular if you want it: a `crossSessionInbound` setting lets a session accept, hold, or refuse incoming messages, `isolatePeerMachines` requires your approval before anything leaves the machine, and administrators can switch the whole feature off org-wide. As 9to5Mac put it, the win is that Claude can warn a session about a breaking change before you even notice it happened. For anyone orchestrating more than one agent at a time, that is the difference between parallel sessions and an actual team. ![Claude]() Claude AI assistant by Anthropic. Write, analyze, code, and create with one of the most capable large language models available. [View Claude →]() Sources [Claude Code docs: Message your other Claude Code sessions](), [@ClaudeDevs announcement on X](), [9to5Mac]() [Read Claude Code sessions can now message each other on Creators Toolbox]() --- Canonical source: https://creatorstoolbox.com/blog/claude-code-session-messaging