About this project

RimZ is a realtime dashboard and control room for agentic coding, built on tmux and Zellij. It puts coding agents in one multiplexer room and routes attention to whichever agent needs it. Each agent gets a live sidebar card showing state, task, model, context health, token stats, dollar cost, and subagent tree. A cockpit line summarizes the fleet, and one click lands in the waiting pane. The tool is a lightweight binary inside existing Zellij or tmux sessions. It supports running official agent CLIs such as Claude Code and Codex unchanged, with hooks reporting activity to the sidebar. Users can launch layouts into isolated Git worktrees, combine models as named teams with roles and handles, and message agents durably with parking, steering, and scheduling options. RimZ provides automation primitives for harness and loop engineering. `rimz agents -p` runs headless supervised agent tasks with exit codes and JSON output for scripts and CI. `rimz loop` schedules runs on calendars, intervals, cron, watchdogs, or event signals. `rimz wait` lets agents arm timers, commands, process watches, or file watches and end their turn instead of sleeping. Dollar budgets cap spending per turn, agent, loop task, room, or provider account. Notification handlers can run custom commands when attention is needed. Remote work is supported through SSH with self-healing reconnection, browser tunneling via ttyd, and optional remote-control bridges for answering prompts from official Claude and ChatGPT mobile apps. On Linux, optional bubblewrap mount views provide sandbox isolation for agent panes. Configuration is file-based under `~/.rimz/`, with TOML config and Markdown definitions for agent profiles, teams, and skills. The project is beta software, heavily used to build itself, and ready for personal daily use while production workflows may prefer waiting for the 1.0 release.