I've been running multiple Claude Code sessions in parallel across different repos and got tired of cmd-tabbing between terminal tabs trying to figure out which one needs me and which one is still working. So I built a little desktop app for it (or I watched it being built as they say these days).
It auto-discovers all running Claude processes and shows what each one is doing — no setup needed. There are a million projects out there that do that but the thing that always felt missing was having the full lifecycle of a code change in one place. You can see an agent go from working, to waiting for approval, to having an open PR with CI checks, review comments, and merge conflicts — all without leaving the dashboard. If a test fails or someone leaves feedback, you see it right there next to the session that wrote the code.
The "Ah yes this code from agent 69 never made it to Github before the release I see" shenanigans are a lot less likely with Claude Control in my experience.
I usually have it fullscreen to the side and everything opens on my main external screen. That way I never get lost in the weeds. And I'm intentionally not implementing any "depth" to the dashboard, opening a detailed session view etc. as that feels like a loss of the "oversight" I'm aiming for.
The trickiest part has been status detection. Claude Code exposes lifecycle hooks that I use as the primary signal, but there are edge cases — a session can look idle when it's actually just between tool calls, or show as working when it's really waiting for input. I've layered on CPU and JSONL heuristics as fallback but it's still not perfect.
My father in law lost his job 2 years ago, completely unrelated to AI, at the age of 63. He had worked this same job since he was 18 and it was his identity. He'd almost greet people with "Hey I'm CEO, my friends and family call me X".
Being fired from his own company completely destroyed him - and not because he's worried about the financial aspect, he doesn't need to worry about that at this point.
I remember watching him go though this and thinking to myself "Jeeze, I'm glad my identity isn't completely tied up with what I do for work".
With AI knocking on the door I'm surprised how much my identity and perceived self-worth is actually tied to being a "good" developer. But it's more of a slow burner than what my father in law got. So I at least have some time to mentally prepare for my new reality.
Makes you wonder with some of the dinosaurs in office if a big reason why they cling on is that internal mythmaking. Presumably everyone knows they can just up and kick back by some body of water doing whatever they'd like with their time, should they ever want to step down. And yet they don't. They prefer their busy, stressful, life of importance.
Me personally I'm just not wired like that. I work hard only to get to that payoff of being lazy and carefree later. I don't actually like being in this stressful state. It bewilders me frankly that some seemingly are addicted to it. Maybe they literally are, considering flight or flight response is a real measured altering of neurotransmitters.
The core of what we were doing (Writing code) dictated the core of our tool kit (IDE).
Now that we're not writing code anymore it's very exciting to see how this unfolds in the tool kit.
Before you bury me for that "Not writing code anymore" statement my current workflow is literally just talking to Claude via Wispr Flow (mic) whether it's describing what we need to implement or reviewing what Claude Code implemented.
> The core of what we were doing (Writing code) dictated the core of our tool kit (IDE).
>
> Now that we're not writing code anymore it's very exciting to see how this unfolds in the tool kit.
So maybe the text area in your IDE becomes read-only. Even when not actively debugging, you still need to read code and efficiently browse through it as you review it. Because you always review code, don't you? Don't you??
Hook Hero is a live mission control for all your Claude Code sessions — showing you at a glance which agents are running, how many sessions you've had today, time spent per project, cost per session, message counts (interactive vs CLI), tool usage, and the real-time status of every active
and completed run.
I've been running multiple Claude Code sessions in parallel across different repos and got tired of cmd-tabbing between terminal tabs trying to figure out which one needs me and which one is still working. So I built a little desktop app for it (or I watched it being built as they say these days).
It auto-discovers all running Claude processes and shows what each one is doing — no setup needed. There are a million projects out there that do that but the thing that always felt missing was having the full lifecycle of a code change in one place. You can see an agent go from working, to waiting for approval, to having an open PR with CI checks, review comments, and merge conflicts — all without leaving the dashboard. If a test fails or someone leaves feedback, you see it right there next to the session that wrote the code.
The "Ah yes this code from agent 69 never made it to Github before the release I see" shenanigans are a lot less likely with Claude Control in my experience.
I usually have it fullscreen to the side and everything opens on my main external screen. That way I never get lost in the weeds. And I'm intentionally not implementing any "depth" to the dashboard, opening a detailed session view etc. as that feels like a loss of the "oversight" I'm aiming for.
The trickiest part has been status detection. Claude Code exposes lifecycle hooks that I use as the primary signal, but there are edge cases — a session can look idle when it's actually just between tool calls, or show as working when it's really waiting for input. I've layered on CPU and JSONL heuristics as fallback but it's still not perfect.
If anyone has clever ideas here I'm all ears.
reply