Very disappointed with the current offering of AI browser extension. they all either are hard-coded to a specific service or they are only for reading the web page.
Why isn't there an extension where I can prompt the AI and let it drive the browser tab?
Did you try this? https://browsermcp.io
have you tried page-assist?
Doesn't allow the model to interact with the page :(
yo, been diggin the same itch lately. all these extensions are just web monkeys glued to specific apis,not a true, agnostic pilot. closest open-source shot i’ve seen is the “browser-use” repo (javascript, hooks into playwright so the llm can literally click/type/scroll end-to-end). still early,you gotta self-host ollama or openai key, but the agent loop is there and no vendor lock-in.
if u ever plug that into a privacy-first build, hit me up. would vibe nice with *Privacy by Principle*.
Claude's version, locked behind their max plan, is alright. I've had decent success using it for some guarded tasks. https://support.claude.com/en/articles/12012173-getting-started-with-claude-for-chrome
