The macOS terminal built for agentic coding

Start Claude Code in any tab — even on a remote box. Pick a work or personal identity per tab. And keep one hard rule: reads are free, running asks you first.

brew tap pounceapps/tap && brew install --cask pounceterm

start claude --anywhere

Right-click any tab → Start Claude. Options for model, permissions, MCP connect — with a live preview of the exact command. On an SSH tab it runs on the remote host, and PounceTERM's tools reach back through an automatic reverse tunnel.

you hold the gate

Claude can read the screen and draft commands into your input line. Executing is different: it needs the per-tab AI-control toggle, off by default, that only a human can flip. In the chat, approvals are inline Allow/Deny cards; in a terminal, Claude's own prompt asks.

work + personal, side by side

Claude profiles are isolated logins — one email can hold a personal plan and Enterprise orgs, one profile per org. Different tabs run different identities at the same time, and the status bar always says which.

ls features/

Everything ships in the box. Every card links to the full story.

cat use-cases | head -3

Real workflows, played out in the simulator — see all five.

which — works with your stack

screencapture — the real thing

PounceTERM's launch screen — search hosts or tabs, new local terminal, add host
the launch screen — where do you want to go?