Lamoom is a marketplace of agent apps that run inside your own Claude - not on our servers. Install one and it works in your files, with your data. Every run is yours to see. Publish your own loops and charge per run or monthly. $20 free credit to start.
No reviews yetBe the first to leave a review for Lamoom
Maker
📌
Hi Product Hunt 👋
I'm Kate. I built Lamoom alone, in San Francisco, mostly on AWS Lambdas.
Every person I know who builds with Claude has the same folder: three or four agent workflows that work really well and that nobody else will ever use. There
was no way to hand them to someone. No way to charge for them.
Lamoom is that missing piece. Apps run inside your own Claude - your files, your
data, your connectors. Publishers set their own price. Eight apps are live today; publishing is open to anyone.
So Lamoom is a marketplace for them. Two things make it different:
1. Apps run inside YOUR Claude. Your files, your data, your API, your connectors. I can see that a run happened. I can't see what's in it.
2. Every app is a loop with judges. It doesn't return the first draft - it scores its own output and iterates until it passes the bar. That's why the news digest doesn't email you something boring.
And please try it, run news collection for your industry:
https://lamoom.com/app/claude-ne...
There are 8 apps live today, built by me. Claude News - is an app you should try. Publishing is open now: you keep the loop, you set the price, per run or monthly.
$20 free credit when you sign up, which is a lot of runs.
I spent 13 years as an engineer, ex-Amazon, where "mechanism" meant a
process that keeps working after the person who cared about it leaves. Lamoom is
Lambda + mechanism. That's literally where the name comes from.
I'll be here all day. Tell me what's in your folder - I want to know what people
would publish first.
the "loop with judges" idea is the part I'd want to poke at, since it's basically the same shape as the eval problem everyone's fighting with agents right now. if the judge scoring the output is itself an LLM, what stops a loop from converging on an answer that's confidently wrong but happens to satisfy its own judge's bar? that's different from just returning a bad first draft, it's a wrong answer that's been polished until it looks right. do publishers get any visibility into how often their judge is overruling itself vs just rubber-stamping the first pass
Gal Dayan's judge-loop question is one I'd want answered too, but the angle I keep coming back to is the marketplace side of it. a publisher tunes their loop against whatever Claude does today, then Anthropic ships a model update and the underlying behavior shifts under everyone at once. if a published app starts behaving worse after that, who eats it - does the publisher get a heads up to retest, or do buyers just quietly get a worse loop for the same price until someone notices and complains?
Report
Ran a few loops in my own Claude and the data staying local thing is honestly the part that sold me. Setup took like two minutes, and I could actually see every run which felt surprisingly transparent for this kind of tool.
Report
Maker
@formfl Yes, it was the goal - to make a setup saved in between, and you see how it run out of tokens without telling any result
Report
@kate_yanchenko This is super cool, a marketplace for claude skills and loops. I just starting exploring. One suggestion, its using he google auth url and looks sus. You can prompt the AI agent to configure it and just have sign in to Lamooom. That will look very clean.
Report
Maker
@roopesh_donde love your suggestion, thanks for helping to improve the product!
Agentplace
is it only for Claude, or can I use it for codex/chatgpt as well?
@vlad_yanch yes, it works with chatgpt
Dial
the "loop with judges" idea is the part I'd want to poke at, since it's basically the same shape as the eval problem everyone's fighting with agents right now. if the judge scoring the output is itself an LLM, what stops a loop from converging on an answer that's confidently wrong but happens to satisfy its own judge's bar? that's different from just returning a bad first draft, it's a wrong answer that's been polished until it looks right. do publishers get any visibility into how often their judge is overruling itself vs just rubber-stamping the first pass
Dial
Gal Dayan's judge-loop question is one I'd want answered too, but the angle I keep coming back to is the marketplace side of it. a publisher tunes their loop against whatever Claude does today, then Anthropic ships a model update and the underlying behavior shifts under everyone at once. if a published app starts behaving worse after that, who eats it - does the publisher get a heads up to retest, or do buyers just quietly get a worse loop for the same price until someone notices and complains?
Ran a few loops in my own Claude and the data staying local thing is honestly the part that sold me. Setup took like two minutes, and I could actually see every run which felt surprisingly transparent for this kind of tool.
@formfl Yes, it was the goal - to make a setup saved in between, and you see how it run out of tokens without telling any result
@kate_yanchenko This is super cool, a marketplace for claude skills and loops. I just starting exploring.
One suggestion, its using he google auth url and looks sus. You can prompt the AI agent to configure it and just have sign in to Lamooom. That will look very clean.
@roopesh_donde love your suggestion, thanks for helping to improve the product!
Morejust: store
Looks really cool! I wonder if there’s any way to filter more useful and less useful apps?