Blog
Harjas Singh
It's 2pm on a Tuesday. You have four Claude Code sessions running. One is refactoring your auth middleware. Another is writing integration tests for the payments flow. A third is doing a database migration you've been putting off for weeks. The fourth is exploring a new API you need to integrate by Friday.
Each one is humming along. The problem is you have no idea which ones are still humming and which ones stopped ten minutes ago.
Every time you alt-tab to check on a Claude session, you pay a tax. It's not just the 2-3 seconds to switch windows. It's the 20-30 seconds to re-orient: what was this session doing? Is it still running? Did it ask me something? Where was I before I switched?
That adds up fast. If you're checking on sessions roughly 20 times per hour — and that's conservative when you have four running — you're burning 6-10 minutes every hour just on status checks. Not doing work. Not reviewing output. Just checking if something needs your attention.
Over a full workday, that's 50-80 minutes gone. Not to mention the cognitive load of holding four different problem contexts in your head simultaneously.
The terminal is great for one session. It's fine for two. At three, you start losing track. At four or more, it falls apart.
Terminal tabs don't show status. You can't see if Claude is running, waiting for approval, or finished without clicking into each tab. Multiply that by the number of sessions and you're just polling — manually cycling through tabs every few minutes to see if anything changed.
Some developers try splitting their terminal into panes. Now you can see more at once, but each pane is tiny and you're still scanning text to figure out status. Your eyes do the work your tools should be doing.
The fundamental issue: terminals were designed for one active process at a time. Claude Code sessions are long-running, asynchronous, and need periodic attention. Those two models clash.
Your MacBook has a piece of always-visible screen real estate that most apps ignore: the notch. It sits at the top of your screen, visible no matter what app you're in, no matter which Space you're on.
Claude Pulse turns that notch into a session dashboard. Each active session gets a small breathing dot. The color tells you everything at a glance:
One quick look at the notch. No alt-tab. No window switching. No context loss. You know immediately which sessions need you and which ones are fine on their own.
The notch expands on hover to show session details — what each session is working on, how long it's been running, what it's waiting for. But most of the time you don't need that. The dots are enough.
When a session needs approval, the amber dot tells you. But Claude Pulse goes further than just showing status.
Hit Cmd+Y from anywhere. The notch expands. Your sessions are listed with number keys — hit 1, 2, or 3 to select the session that needs attention. Hit Enter to approve. The whole interaction takes under two seconds and you never leave your current window.
Claude Pulse is also focus-aware. It knows which terminal tab or tmux pane you're looking at. When you hover the notch, the session matching your current terminal is highlighted first. No hunting through a list to find the right one.
For tmux users running multiple sessions in a single terminal, this is a big deal. Claude Pulse detects individual tmux panes and maps them to the correct session. It works with up to six simultaneous sessions across any combination of terminals, tabs, and tmux layouts.
Here's my typical workflow now. I kick off three or four Claude sessions in the morning. I work on design or review PRs in another window. Occasionally I glance at the notch — all green dots, everything's running.
An amber dot appears. I hit Cmd+Y, see that the migration session wants to run a destructive query, review it, approve it. Back to what I was doing in under five seconds.
At the end of a session, I might have approved 15-20 operations across four sessions without alt-tabbing once. The sessions that only needed safe operations (reading files, running tests) handled themselves. I only got pulled in when something actually needed my judgment.
The features page has the full list of what Claude Pulse tracks. The comparison table shows how it stacks up against checking terminals manually or using other monitoring tools.
The best session management is the kind you don't think about. You shouldn't be spending mental cycles on "which session was doing the auth refactor again?" or "did I check on the test runner recently?"
The goal isn't to manage sessions. It's to forget about them until they actually need you. The notch gives you that. Ambient awareness without active attention. You notice the amber dot the same way you notice a notification badge — peripherally, without breaking flow.
Four sessions, six sessions, it doesn't matter. They run. You work. When one needs you, you'll know.
More from the workshop
Running Claude Code in tmux means juggling approvals across panes you can't see. Here's why standard notification approaches break down — and how Claude Pulse routes approvals to the right pane by PID.
The first prototype had a login button. I tore it out. Here's why Claude Pulse reads the CLI's existing Keychain credentials instead of asking you to authenticate twice.
Every missed approval costs ~10 minutes. With multiple sessions, that's hours per day. Here's the math and how safety-aware batch approval fixes it.