Has anyone gotten this Claw to work properly?

[Onair-santa]

Has anyone gotten this Claw to work properly?
Looks more like a fake.
The docs were written by the AI ​​agent.
The docs contain command errors.
The wizard contains script errors; I can't install the databases and agent properly.
The app doesn't launch or work on Windows 11.
Correct me if I'm wrong, but this isn't a working version with a pretty picture on the home page.

[seohyunjun]

It works but Doesn't better than openclaw

[CorellisOrg]

it works but you need to be realistic about what it gives you vs vanilla openclaw. the multi-tenant part is genuinely useful if you're managing agents for multiple users/teams. postgres backend is more production-friendly than file-based state.
where it gets tricky (and this isn't goclaw-specific, same with openclaw): once you have more than 3-4 agents, the coordination problem dominates. agents duplicate work, context windows fill up with stuff another agent already figured out, and you end up manually dispatching tasks because there's no automatic goal breakdown.
we hit this running 8 agents and ended up building a governance layer that sits on top of whatever runtime you use. goal decomposition handles the "what should each agent work on" part, and fleet learning handles the "agent B shouldn't repeat what agent A already discovered" part. works with openclaw, should work with goclaw too since the governance is runtime-agnostic.
wrote it up at https://github.com/CorellisOrg/corellis — might be relevant to the context length issue (#613) too since shared memory means agents don't need to carry everything in their own context.

[seohyunjun]

I was dissatisfied with the fixed length of Cotext or the use of too many tokens as output. It's the same issue as you said, and it's fatal in the open model. The agent that needs to use multi-turns immediately stops because of this problem.