Before I Get Frustrated, Is This Really Worth My Time?

[directorboint-arch]

OpenClaw was hyped by YouTube influencers and it didn't' deliver what they promised and came with security flaws. I've unsubscribed to all of them for wasting my time. And nemoclaw seemed better at first and then the nightmare began...

Many users are reporting that the OpenShell Sandbox networking is failing or that the Kubernetes (k3s) namespaces required for the gateway are not initializing properly. Because NemoClaw tries to wrap a complex security layer around an agent, if one part of that stack (like Docker cgroups or the K8s pod) fails, the whole "one-command" setup breaks. NemoClaw and OpenShell are essentially trying to build a high-security vault around a very "leaky" engine (OpenClaw), and right now, the vault door keeps locking the owner out.

What problems am I looking at with this codebase?

[shirik]

I can't tell you if / what problem you hit. I got originally my OpenClaw setup running, yet, well, you wrote the story. NanoClaw works very well for me, the only hiccup was Apple containers vs Docker for me (running on OS X). At the moment, I am on docker side and planning to get back to Apple container. The switch / configuration changes are comfortable for me.

The only downside - and there is another thread about it - is that Claude configures it (depends on you trust to Claude) and with fork, I am not super happy to have my public repo.

Yet, overall, I am happy with this project.

[directorboint-arch]

Thanx for your thoughts. I think the problem with Nemoclaw is that it doesn't work well on WSL with docker. Seems to run well on OSX and native linux. I might have to dig out an old laptop. But, last time I did that, OpenClaw made my system unbootable and I lost two weeks of work. sigh

[shirik]

I feel your pain. My first run of OpenClaw was my MacMini Late2012 (with i7 & 16G). I had to go through OpenCore legacy patcher and than I ended up that "it is unusable after sleep" when OpenClaw runs there. So I moved to Linux server with Linux Mint (I used to be an excited desktop user of Linux Mint before moved to MacBook). And you know the story from there.

If I was in your current position, I would probably wait give the change in Antrophic usage for 3rd party systems (technically NanoClaw now fits there). And I believe it depends on what you want to do. With all the recent changes from Antrophic, I run a few things natively in Claude itself.

Therefore, I still see a value in NanoClaw to automate the advanced flows and management of backlogs, servers, producing some artefacts for you. Yet, I believe the "use cases" are getting smaller every day with Antrophic almost reading my mind.

The other option is if you want to use if primarily with "offline models" and I am not certain that NanoClaw is the good fit here (though I believe it might - I haven't tried) as I believe Clause is first class citizen here.

Btw. I am not trying to encourage or discourage you from NanoClaw use and it is solely my personal opinion / experience so feel free to challenge any of it.

[glifocat]

@shirik v2 includes a deterministic bash script for setup instead of through Claude Code and supports Ollama, Opencode and Codex now. Would love for you to try it and let us know what you think.

[directorboint-arch]

I'm just using hermes at this point.

[RikyZ90]

Here’s an alternative worth checking out: https://github.com/RikyZ90/ShibaClaw
I'm looking for testers and contributors <3

[directorboint-arch]

That's my next project! I saw that a while ago and I'm looking at getting back to it. I normally mention it when I talk about openclaw or hermes with folks irl