Skip to content

Dashboard ​

One running stack and you remember where everything is. Four running stacks started in parallel by four agents and you don't.

The dashboard at http://lich.localhost:3300/ is a single view of every running lich stack on the machine.

Open it ​

bash
lich dashboard

This auto-starts the daemon if needed (any directory works, you don't have to be in a worktree) and opens the dashboard in your default browser. Pass --no-browser (or set LICH_NO_BROWSER=1) to print the URL instead of opening it.

You can also navigate to it directly: http://lich.localhost:3300/. The *.localhost resolves to loopback on every OS, no DNS setup.

What you see ​

Per running stack:

  • Source worktree: which directory the stack belongs to (and the resolved ${worktree.name} / ${worktree.id}).
  • Services: every compose service and owned process, with status (starting / ready / failed / stopped) and uptime.
  • Friendly URLs: clickable <service>.<worktree>.lich.localhost:3300 links, one per service.
  • Logs: tail any service inline; full per-service log files are at ~/.lich/stacks/<id>/logs/<service>.log.
  • Restart / stop: per-service or whole-stack restart, stop the whole stack.

Why this matters for parallel agents ​

The dashboard is what makes N-agents-in-N-worktrees observable. Without it:

  • You forget which worktrees have running stacks.
  • A failed service in agent 3's stack looks identical to a failed service in agent 7's.
  • Tailing logs means remembering the full state-dir path of each stack.

With the dashboard, all of that goes through one URL. Agents can deep-link to specific services' logs (http://lich.localhost:3300/stacks/<id>/logs/<service>) so when an agent says "the api in worktree X is failing," you have the URL to verify.

Auto-launches on first stack ​

The daemon (and therefore the dashboard) autostarts the first time any lich command needs it lich up, lich dashboard, lich urls, etc.

See also ​

Released under the MIT License.