I really love this project a lot, but there are a few issues that came up during my heavy usage today

[luck7x]

First of all, big praise — it's really much more stable than openclaw.

But there are a few things below that give users a bad experience:

I don't know yet if there's something like openclaw's /verbose to show the agent's operations [When I'm chatting, I don't even know what it's doing or whether it has crashed — Codex executions usually take quite a long time]

Regarding the queue issue, I observed it for an afternoon and found that processes often fail because of this

[josephbolus]

I also would want more visibility on what each agent is doing. But if you have the claude-mem plugin installed you can actually see this activity by going to http://localhost:37777/ which is pretty neat!

I asked my agents to build a "mission control" dashboard and pull live data from the claude-mem worker API in http://localhost:37777/stream

[jlia0]

relevant PRs:

#113
#106

[jlia0]

I also would want more visibility on what each agent is doing. But if you have the claude-mem plugin installed you can actually see this activity by going to http://localhost:37777/ which is pretty neat!

I asked my agents to build a "mission control" dashboard and pull live data from the claude-mem worker API in http://localhost:37777/stream

The mission control dashboard looks very cool, are those metrics / protocols legit?