How to keep track of Claude Code, Codex and Cursor

Running three agents at once feels like a superpower until you notice where your time goes: checking on them. Which terminal was the auth fix in? Did Cursor finish? Has Claude been waiting on a permission prompt for ten minutes?

The real cost of parallel agents

Agents are fast. The slow part is you finding out they need you. Every check is a hop: find the window, find the tab, read the last few lines, go back. Multiply by three agents and a workday.

One list, sorted by who needs you

In Hopper, press ⌥Space from anywhere. Every agent is in one list, grouped by status:

Status	Means
Needs You	Stopped to ask you something, like permission to run a command
Done	Finished since you last looked
Working	Busy right now. Leave it be
Idle	Nothing new to see

Press ↵ and you’re in that exact session, in the terminal, editor or app it’s running in. Once you’ve seen a Done agent, it moves to Idle on its own.

One key to whoever’s waited longest

Don’t even open the list. Next Agent, ⌥⌘], jumps straight to the agent that needs you, then the ones that finished. Press it again for the next. It’s a queue you work through with one key.

Find the session from last week

Past Claude Code and Codex sessions are in search too. Press ⌘Space, type what it was about (“rate limiting”, “teams migration”), and open it again.

No hooks, nothing installed

Hopper doesn’t install hooks or plugins into your agents. It reads each agent’s own files on your Mac. Claude Code and Claude report when they’re waiting on you; Codex and Cursor show Working, Done and Idle.

It works with Claude Code, the Claude app, Codex and Cursor, plus agents running in herdr.

Try it

Hopper is in private testing for macOS 26. Join the waitlist.