An interesting plugin found on GitHub: Nanobrowser
This is a Chrome extension that lets you control the browser to work for you using natural language once installed.
It now has over 10,000 stars.
For example:
I asked it to check the top three papers on Hugging Face,
read the titles, likes, and abstracts, then sort them by likes.
It automatically navigates webpages, reads content, organizes information,
without me doing anything, and finally sends me the results.
Watching it work is quite interesting,
it tags each element on the page and then clicks and operates by itself.
Technically, it uses two agents:
Planner is responsible for understanding tasks and breaking down steps
Navigator handles the specific webpage operations.
If there’s a problem, Planner dynamically adjusts the strategy.
As for cost, I used the DeepSeek API,
and this one task cost me just 0.1 RMB… really just 0.1 RMB.
Configuration is simple, just install the extension and set up an API key to use.
For scenarios involving repetitive web operations,
this thing is quite practical.