Agent Browser, Port Forwarding and ZeroClaw
05 May 2026
I have been continuing my experiments with ZeroClaw and its browser
integration. Most of my goals with AI are explainable AI. I want to be able to
see the tool calls that it makes to answer my questions. I want to know if the
query was properly grounded. To answer some of these questions I enabled otel
so I can see the tool calls in grafana, I also encourage the agent to use tools
in the prompts. However, using the browser seems to be a challenge for the
agent on a headless server (how do rpa tools do this stuff anyways?). I also
want to ensure that my use of AI is as private as possible for me. The most
important part of any summarizer/explainer is how did it come to this
conclusion.