Founder shared 2026-08-31 asking "do we have a need for this?" Verdict: skip for now, re-look on trigger.
What it is
Runtime that keeps coding agents (Claude Code, Codex, 21+ CLIs) running in persistent background sessions that survive disconnects and restarts. Status monitoring (working/blocked/idle), multi-agent coordination via shared socket API, cross-platform. Open source (Apache 2.0), YC-backed, v0.8.x as of filing. Install = curl|sh.
Why we skipped
The RDCO always-on stack ([[project_channels_agent_setup]]: Mac Mini + tmux claude-channels + LaunchAgent + 4am restart + fleet manifest) already provides the core value: sessions never depend on a client connection, reconnect = tmux attach. Adopting would swap battle-tested tmux for an early-stage dependency at the base of the whole stack — real risk, cosmetic gain. Any future install goes through the standing pre-install security review ([[feedback_mcp_install_security_review_default]]).
What it has that we lack
- Cross-agent status dashboard (working/blocked/idle) — we approximate with
~/.claude/state/fleet-status/files + logs - Cross-machine session migration — we are single-box by design
Re-look triggers
- We add a second always-on machine
- Fleet grows past ~5 concurrent agents and status visibility becomes a real pain