Herder keeps familiar terminal-multiplexer features such as workspaces, tabs, split panes and persistent sessions, then adds a sidebar that understands supported coding agentsAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository.. Instead of manually checking every pane, a developer can see which agent is still working, waiting for approval, idle or finished but not yet reviewed.
State detectionAI agent state detection determines whether an agent is available, running, waiting, blocked, failed, disconnected, or complete. combines direct lifecycle integrations with conservative screen matching for agents that do not expose hooks. Optional notifications surface blocked or completed background work, while unsupported tools still run as ordinary terminal processes without rich status tracking.
The local CLI and socket API make the multiplexer an orchestration layerAgent orchestration coordinates AI agents, tools, people, tasks, state, and control flow so a larger workflow reaches a verified outcome. rather than just a dashboard. One agent can split a pane, launch another agentAgent-to-agent communication lets AI agents exchange messages, task state or other information with one another., submit review instructions, wait for a target state and read the result back, enabling structured multi-agent workflowsA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks. without brittle key-sending scripts.
A client-server design keeps terminal sessions running across disconnects and can restore supported agent conversations after a restart. Remote mode moves the server to another machine while preserving the local interface, although pre-1.0 protocol changes, heuristic detection, incomplete Windows support and nested multiplexers remain practical limitations.
Watch on YouTube
