herdr

herdr is a terminal-native runtime and multiplexer for coding agents. A background server owns terminal sessions so agents can keep running while clients detach, and users can reattach from another terminal or over SSH. It also labels agent panes by attention state—working, blocked or idle/done—so developers can scan multiple concurrent sessions without manually checking every terminal.¹

The runtime is designed to orchestrate existing coding agents rather than replace them. Its CLI and local socket API let software create panes, run commands, read output and coordinate work between sessions. That makes herdr useful as infrastructure around tools such as Claude Code, Codex and other terminal agents, while preserving each agent's native terminal interface.¹

The current canonical `herdrdev/herdr` repository is public and licensed under Apache License 2.0. It can be built from source as a Rust project, and the repository points to enterprise/partnership contact separately. AiToolMap therefore treats the core runtime as open-source/free-access software rather than inferring a proprietary seat-price model.¹

Some third-party forks and older indexed copies still show previous AGPL/commercial-license language. Those surfaces are not the current canonical repository and are excluded from the product facts used in this review.

The canonical GitHub surface currently shows roughly 36,000 stars and 2,600 forks. Those are useful adoption signals, not measurements of reliability, agent-detection accuracy or productivity. AiToolMap has not hands-on tested session persistence, restart recovery, state detection or socket/API behavior.¹

Y Combinator lists herdr as an active Fall 2026 company and describes the same open agent-runtime architecture. YC also carries company-supplied adoption statements about stars, downloads and plugin growth; those are not treated as independent efficacy evidence.²

The fixed 50-source independent panel found no current exact-product third-party benchmark, substantive independent review or eligible user-rating corpus. Forty-three sources returned no relevant exact-product result and seven canonical domains were crawler-inaccessible. Missing external evidence is neutral.

herdr is best for developers and teams already running multiple coding agents in terminals and feeling coordination overhead: lost sessions, agents waiting unnoticed, too many SSH/tmux contexts or a need for agents to spawn and inspect other terminal work programmatically.¹

It is less compelling for users who run a single agent occasionally or prefer a fully managed graphical cloud environment. Teams depending on it for long-running production workflows should validate restart behavior, agent integrations, socket compatibility and security boundaries against their own environment because the runtime and ecosystem are still evolving.