Hey IH. I'm launching Conan on Product Hunt today and wanted to share it here too.
https://www.producthunt.com/products/conan
A native Mac app that sits beside Claude Code and shows what it's doing live: a streaming timeline of every tool call and skill, a context-window gauge, a usage pulse, and a skills/MCP browser. All on-device, no telemetry. Would love your feedback.
I'd be careful treating this as an observability tool too quickly.
The interesting question may not be what Conan shows developers.
It may be what decision developers become more confident making because it exists.
Those sound similar, but they can lead to very different conclusions about positioning, adoption, and what feedback matters.
I wouldn't make that call casually from launch feedback alone.
Really good distinction, and an uncomfortable one in the best way. Observability is the easy story to tell. "What decision does it make you confident enough to make" is the harder, truer one.
So I'm taking your advice fairly literally: not making that call from launch feedback. Noting it, and watching what people actually do with it over the next few months. Appreciate you thinking about it this carefully.
Possibly.
The reason I'd still be careful is that some of the most expensive decisions don't happen when the feedback arrives.
They happen when founders decide what the feedback means.
That's one of those things I'd want confidence in before too much gets built around it.
I wouldn't try to unpack that properly in a thread.
If you're curious, drop your email and I'll put together the tighter version.
The UI looks great. I especially like the focus on local-first and no telemetry.
One feature I'd personally find useful is a condensed activity summary ("What did the agent actually do?") alongside the live timeline, especially for long-running sessions
Thanks! Local-first + no telemetry is the whole point, glad it shows.
The "what did the agent actually do?" digest is a great call — the live Timeline captures it row-by-row today, but a condensed run summary (files touched, commands, skills, outcome) is a different and better thing for long sessions.
Definitely on the roadmap. What would you want it to surface first?
for me, the most valuable summary would be like these
-Files touched
-Commands run
-Skills used
-a one-line outcome summary
something like "Updated authentication logic across 5 files and fixed failing tests"
That would make it much easier to jump back into long sessions
as someone who lives in claude code all day, the context-window gauge is the part i'd open this for, watching it silently fill until things degrade is the current pain. the streaming timeline i'd probably glance at once and forget tbh. the on-device no-telemetry bit matters more than you might think for the work-laptop crowd who can't run random tools. my one worry is it becomes a thing i check instead of a thing that tells me when to care, idk if a passive alert beats a live dashboard here
A bit more on the why: I basically live in Claude Code, but mid-session I could never tell what it was actually doing, which tools it fired, how full the context was, or which skill it grabbed. The terminal just scrolls and you end up trusting it blind.
So Conan is the thing I wanted on my second screen. One detail people have latched onto: it shows you which skill Claude Code actually picked each turn, and even which ones it considered, since the agent doesn't surface that itself.
The bet is that an always-on native HUD you just glance at beats the great-but-CLI tools already out there. Genuinely curious whether that resonates with folks here, and if you've shipped a native Mac app, I'd love to hear what worked for distribution.
https://conan.sh