Message Your Other Claude Code Sessions: Cross-Session Messaging

Claude Code v2.1.224+ (macOS, Linux) now lets Claude deliver messages between your independent sessions. When a change in one session breaks what another is building on, Claude can warn that session before you notice. When one session settles a question another is blocked on, Claude can send the answer across. No configuration needed — it's on by default.
How it works
Claude uses two tools: ListAgents to discover which sessions it can reach, and SendMessage to deliver a message to one by name. The same tool also works for subagents and agent team teammates, but this page focuses on independent sessions.
When to use it
- Hand over a finding: one session discovers a breaking change or makes a decision — Claude summarizes it for the affected session.
- Coordinate parallel worktrees: sessions working the same repo in separate worktrees can be told what landed.
- Get status from long-running work: migrations or test runs can report back to the session you're watching.
- Reply across machines: answer a message that arrived from another machine — but only reply, not initiate.
Use this between sessions you start and steer yourself. For other multi-session workflows, Claude has dedicated features:
- Resume a session to continue one conversation in another terminal.
- Agent teams for coordinated, Claude-spawned sessions.
- Agent view to watch and steer many sessions from one place.
- Remote Control to steer a session from your phone.
- Channels to push external events (CI results, chat) into a session.
Sending a message
You don't call the tools directly — Claude decides when to send, or you can prompt it. For example:
Ask the session running in my other terminal whether the migration finishedClaude writes the actual message, so your prompt can leave content to Claude. For a summary, you might say:
Explain what we just did to the session working on the payments APIDelivery and controls
The receiving Claude reads the message between tool calls during an active turn, so running tools aren't interrupted. If the session is idle, Claude Code starts a new turn. Delivery is subject to inbound controls, with three outcomes:
- Delivered: passed to Claude.
- Held: set aside until you approve or settings change.
- Refused: dropped.
Delivered messages count toward usage like a normal prompt. Permissions remain per-session. Cross-machine, only replies are possible.
📖 Read the full source: HN AI Agents
👀 See Also
LTM: A JSON Protocol for Portable Agent Memory Across Models and Machines
LTM is a JSON protocol (Core Memory Packet) plus CLI/server for persisting agent context—dead ends, constraints, next steps—across models, editors, and machines. Packets are 2-5 KB, redact secrets, and support MCP.

Indie Developer Unveils 'Ideanator' CLI Tool for Structuring Vague Ideas with Local LLMs
Ideanator is a CLI tool designed by a self-taught 19-year-old developer using local LLMs like Ollama/MLX. It structures vague ideas into well-defined concepts, completely offline.

JobPilot: Claude Code Plugin for Automated Job Applications
JobPilot is a Claude Code plugin that automates job searching and application processes using Playwright browser automation. It includes commands for searching job boards, auto-filling applications, generating cover letters, and tracking application statistics.

AgentMail Founder Details Agent-Native Onboarding After OpenClaw Exposed CAPTCHA Block
AgentMail, an email API for AI agents, rebuilt its onboarding flow after its own OpenClaw agent failed at a Cloudflare CAPTCHA. The new system offers a single REST endpoint for programmatic account creation while keeping humans in the loop for verification.