News for hackers. Stuff that matters.

Show HN: Agent Hand – Tmux session manager for AI coding agents (Rust)

https://weykon.github.io/agent-hand/

I was juggling 5+ Claude Code instances simultaneously. Four terminal windows, each with 3-4 tmux panes. "Which Claude is working on which task?" Wasting 10+ minutes just finding the right session.

Built Agent Hand, a Rust rewrite of agent-deck with:

- Ctrl+N jumps to the most urgent session (waiting → ready → running) - Visual status from PTY output patterns (regex, no polling) - Sub-millisecond fuzzy search (Ctrl+G) - Dedicated tmux server (isolated from your default)

The WAITING state (!) is detected from prompt patterns like "Do you want to proceed?" The READY state () uses a TTL—if it was running recently and now idle, it's probably done.

MIT licensed. Would appreciate feedback on the PTY monitoring approach and any edge cases I'm missing.

https://github.com/weykon/agent-hand https://weykon.github.io/agent-hand/

( 2 comments | on HN )
Re:Show HN: Agent Hand – Tmux session manager for AI coding ... by jlongo78 on Sunday March 01, 2026 @02:00PM (#47209585)
Rust is a solid choice for tmux session management - the process supervision and PTY handling really benefit from the ownership model. One pattern worth considering: storing session metadata separately from the tmux server itself lets you replay output and search history even after sessions die. Also, multi-user isolation gets tricky fast if agents are running shell commands - worth thinking about per-user socket namespacing early rather than retrofitting it later.
Re:Show HN: Agent Hand – Tmux session manager for AI coding ... by prateekk77 on Monday March 02, 2026 @08:17AM (#47217609)
Nice. PTY pattern matching is the right call for state detection.

I run 16-30 Claude Code sessions in parallel using an orchestrator I built (ComposioHQ/agent-orchestrator on GitHub). Session management is half the problem. The bigger problem is managing your own context around everything that's going on. Which agent is working on what, what's blocked, what just merged, what needs rebasing. The orchestrator is really an optimised human-in-the-loop process. It keeps me oriented so I can make decisions fast instead of spending 10 minutes figuring out what state everything is in.

Have you hit that yet or is this mostly focused on the monitoring side?