Claude Code Sessions Can Now Talk to Each Other Across Terminals

Claude Code Sessions Can Now Talk to Each Other Across Terminals

Agentic AI

Claude Code sessions on the same machine can now send messages to each other in real time — and when the sessions span different computers, they route through Anthropic’s servers. The practical gap this closes is not a minor convenience feature. It is the coordination layer that multi-agent coding workflows have been patching around.

What Changed

Cross-session messaging works differently depending on the deployment context. On a single machine (macOS or Linux), sessions communicate locally and support bidirectional messaging — one session can ask another a question and wait for the response. Across different computers, messages route through Anthropic’s infrastructure and currently support responses only, not full bidirectional queries. The feature is not available on Amazon Bedrock, Google Cloud Agent Platform, or Microsoft Foundry.

Administrators can restrict the feature via settings, giving teams that are cautious about inter-session context sharing a governance control at the config level.

Why It Matters for Builders

The barrier this removes is real. Long-running agentic workflows — a test suite running in one terminal while a refactor runs in another, a documentation agent that needs to know what the coding agent just changed — have historically required the developer to act as the message-passing layer. Copy the output from terminal A. Paste it into terminal B. Describe what just happened. The Claude Code channels feature moved Claude Code toward always-on awareness via external webhooks; cross-session messaging moves the coordination inside the development environment itself.

The distinction matters for how complex tasks are decomposed. When sessions can coordinate autonomously, a task like “refactor this module and update all tests that reference it” can be split into two agents working in parallel, with the test agent receiving context updates as the refactor agent makes changes. Without messaging, that decomposition requires constant human handoffs or a shared file-based state hack that defeats the purpose.

The remote routing path via Anthropic’s servers is the more significant unlock for teams. Local coordination was achievable with clever scripting; cross-machine coordination between a developer laptop session and a CI environment session was not. That path is now native.

The So What

Cross-session messaging is the primitive that makes Claude Code genuinely composable for multi-step agentic workflows. The immediate application is parallel task execution — splitting large refactors, test runs, and documentation updates across sessions that coordinate without human intermediaries. The larger implication is architectural: as sessions can now query each other, the design pattern for complex agentic tasks shifts from a single long-context session to a network of specialized, communicating agents. Builders working on multi-agent production workflows now have a native coordination layer that does not require external orchestration tooling.

Source: The Decoder

Content created with AI assistance and reviewed for accuracy.

💬

Join the conversation

Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.

Join Stack Insiders →

Newsletter coming soon.

Curated digests across AI, biohacking, photography, travel, and more. Be the first to know when we launch.