I tried to build a lightweight browser for agents, like lightpanda/obscura/kindsurf. But I then found out it's a dead end for me and all the others.

The web spec is vast and endless, if you try to build something actually works for modern sites, it will be huge. If you want to keep it lightweight, it won't work for most sites.

Just use chrome-headless-shell or webkit. Don't reinvent the wheels.

Their propaganda has become a footgun. Only the government buys it and limits their access, no target user convinced, what a ironic moment.

This is really a sad moment, I hope the free-theading feature can be kept in the main line. For me, running my library in real threads has actuall revealed some hidden fatal bugs which are hard to reproduce with the GIL>

As for performance, Python 3.x was not even as fast as Python 2.x in the beginning, but with the correct mindset, it's eventually faster.

Because I have to maintain it.

It's mostly only OpenAI, Claude and Gemini may have their unique advantages, but when speaking of models and new paradigm, only OpenAI can do it.

lol what? That’s ridiculous.

> there are just too many features dependent on the full Chrome stack

Do you mind elaborating on what features are missing?

It can be much more easier and realistic with https://github.com/lexiforest/curl-impersonate.

Yes, indeed. I just tried search "Apple", and apple.com is not on the first page.

Let's say you want to use the tool `ssh` and let's imagine ssh has just been released and it's not in the AI agents training set yet.

It won't know in this case how to use the ssh command line tool.

MCP provides a standard way to tell an AI agent how and when to use tools. So if you had an SSH MCP server, you'd simply plug that in, now your AI agent automagically has SSH capabilities.

It's easy and has the word AI in it.

It's infinitely useful for people who's workflows involve LLM agents.

People may do things you have no need for or don't agree with.

Welcome to Planet Earth.

I think it will, it uses requests, and cloudflare blocks traffic from non-browser, e.g. python http clients. It would be better to use something like curl_cffi.