Running multiple Claude Code sessions in parallel should make you faster. In practice, it turns you into a human relay: constant alt-tabbing to copy output from one terminal and paste it into the next, because each instance has no idea the others exist.
A developer built a plugin that fixes this. It adds a direct communication channel between concurrent Claude Code sessions, letting them send messages to each other without a human in the middle.
The use case is straightforward: when you're running parallel coding agents - say, one refactoring a module while another writes tests - they often need context about what the other is doing. Without coordination, you're the bridge. This plugin makes the sessions themselves the bridge.
It works at the plugin level inside Claude Code rather than requiring any changes to the underlying API or model, so it drops into any existing multi-session workflow without extra cost or configuration overhead.
Single-agent Claude Code workflows are well-understood at this point. The harder problem - and the one with more upside - is getting multiple specialized agents to work the same codebase simultaneously without stepping on each other. Most attempts at this either require custom orchestration frameworks or manual context management. A lightweight plugin that handles the messaging layer is a simpler solution than most.
This is a community build, not an Anthropic release, so expect rough edges. But the core idea - that parallel agents need a way to talk to each other, and that the user shouldn't have to be that channel - is the right direction for anyone running serious multi-agent development workflows.