tmux-ide

tmux-ide adds a native layer to existing tmux sessions so the terminal can track agents, show status, and recover from crashes. It is described as teaching the terminal you already use to understand agents, and it can be adopted into a session with a single command. The project emphasizes zero lock-in: if tmux-ide stops, the underlying sessions remain plain tmux.

Its agent-related features include ground-truth status reporting, notifications, and crash-proof restore. The page says Claude Code can be integrated so working, blocked, and done statuses come from the agent’s own lifecycle, while other agents can self-report with a pane option. It also describes a shared bus for agent status, coordination primitives for waiting on status or output, and commands for tasking another agent and reading fleet status in JSON. Continuous snapshots let tmux-ide restore sessions, windows, layouts, working directories, and titles after a tmux server death, and Claude conversations can be resumed from recorded session IDs.

The interface is built as tmux chrome rather than a separate terminal. It includes fleet tabs, live agent glyphs, home and switch triggers, a home cockpit, a sidebar, floating panels for file exploration, git changes, and a config editor, plus an actions menu and a cheat sheet. The UI uses prefix shortcuts and right-click actions, and it supports clickable status glyphs and per-pane border chips. It also works over SSH, and the chrome renders from any client, including a laptop or phone.

tmux-ide is delivered as a command-line tool installable with npm, with commands such as adopt, restore, events, integration install, and worktree. It also supports structured output with --json and optional ide.yml layouts that describe rows, panes, commands, and a sidebar. The page refers to Claude Code, codex, cursor-agent, aider, and other agents, and it notes an MIT link to GitHub in the site header, but it does not state a license there.