The push CLI

Push bundles its own CLI — no separate install. Settings → Workspace → Push CLI and skill links it to ~/.local/bin/push (no sudo) and shows a copy-paste fix if ~/.local/bin isn't on your PATH. The same panel keeps the agent-facing skill current, which is why the agent in your pane already knows these verbs.

push --help is the live authority. The CLI evolves with the app, so treat this page as a map of the families, not a frozen reference.

The verbs you'll actually use

push status        the workspace pulse — what's running, what needs you
push track         turn the session you're in into a tracked issue
push spin-off "…"  bud a follow-up out of this session into its own issue
push doctor        check the install: server, CLI, PATH, runtimes

The families behind them

• Work — push issue … (list, create, comment, complete…), push todo …, push search. Everything the Issues view does, scriptable.
• Sessions & terminal — push session …, push terminal … (open, list, focus…): open and drive tabs, resolve what's running where.
• Workspace — push workspace …, push agent …, push memory …, push schedule … for the standing machinery around your folders.
• Setup & health — push doctor, push update, push app ….

One CLI, two audiences

The point of the design: you and your agent share the same command surface. When you ask Claude Code to "track this session and report back when it's done," it runs the same push track you would — nothing privileged, nothing hidden. That's also why the agents page can promise that everything Push's skill does in your pane is visible in your pane.