herdrdev/herdr

What it solves

Herdr provides a persistent runtime environment for coding agents, solving the problem of losing progress when a terminal session is closed or an SSH connection drops. It allows developers to manage multiple agents across different local and remote machines within a single interface, eliminating the need to manually track which agent is stuck or requires user input.

How it works

Herdr operates as a background server that owns the terminals where agents run. It uses a CLI and socket API that allows agents to interact with the environment—spawning panes, prompting other agents, and detecting when other processes are blocked. It supports existing agent tools (like Claude Code or Cursor) without replacing them, acting as the terminal manager rather than a wrapper.

Who it’s for

Developers building or using AI coding agents who need a stable, persistent environment that supports remote machine management and agent-to-agent coordination.

Highlights

• Session Persistence: Detach from sessions without stopping work; restores layouts and resumes supported agent sessions after restarts.
• Multi-Machine Management: Combine local work and multiple SSH machines into one window with independent reconnects.
• Agent Status Tracking: Clearly marks panes as working, blocked, or idle so users know exactly when an agent needs attention.
• Agent-Native API: Provides a socket API for agents to drive the runtime, allowing them to spawn panes and communicate.
• Lightweight Architecture: Written in Rust as a single binary with no Electron dependency, running directly in the user's existing terminal.