The main advantage of OpenClaw for me was the browser. But Claude Code itself learned to control the browser of my local laptop — remotely, through a reverse SSH tunnel. It reads the screen, clicks, works. The intermediary became unnecessary.
Now the scheme is like this: the browser + local Claude Code lives on the laptop. I connected it to Telegram instead of the old Fedor. I send a command — it flies to the Claude Code console, I see it with my eyes on the screen. The response returns to Telegram.
Result: all the power of console Claude Code — through Telegram + browser control via AI.
I'm not revealing the method of browser control yet. In the next post, I'll show something else — something that shocked even me.
