@lakshmanturlapati
Lakshman Turlapati▲5L2
Applied AI, Voyza · New project · in Agents · 1h ago · edited · Dallas

Shipped: FSB (Full Self Browsing)

I've been building FSB (Full Self Browsing) solo for over a year. It's about 4,000 commits in, which is either dedication or a cry for help, and a couple hundred people use it now.
FSB is a Chrome extension plus an MCP server. Your MCP client is the brain (Claude Code, Cursor, Codex, whatever you run) and FSB gives it 69 browser tools: navigate, click, type, read pages, manage tabs. You don't need an extra API key. It runs in your own signed-in browser, so no fresh cloud browser asks you to log in and pass 2FA all over again. It reads the DOM, selectors and form fields instead of screenshots, which puts most steps at 50 to 200 ms, compared with the 1 to 3 seconds typical of vision agents.

My Grok Bot does all of its browser work through FSB now. I literally set up this Lounge profile by dictating to it over voice while it signed in with GitHub and filled everything out in real time, and it felt amazing. It worked great back when I was running my Claw too. Honestly, after living with it, the screenshot-based options feel a generation behind.

It works on Chrome, Edge, Brave, Arc, anything Chromium.

If your agent needs a site with no API, try it and tell me where it breaks.

github.com/fullselfbrowsing/FSB

FSB (Full Self Browsing) logoFSB (Full Self Browsing)@lakshmanturlapati/fsb · full-selfbrowsing.comI've been building FSB (Full Self Browsing) solo for over a year. It's about 4,000 commits in, which is either dedication or a cry for help, and a couple hundred people use it now. FSB is a Chrome extension plus an MCP server. Your MCP client is the brain (Claude Code, Cursor, Codex, whatever you run) and FSB gives it 69 browser tools: navigate, click, type, read pages, manage tabs. You don't need an extra API key. It runs in your own signed-in browser, so no fresh cloud browser asks you to log in and pass 2FA all over again. It reads the DOM, selectors and form fields instead of screenshots, which puts most steps at 50 to 200 ms, compared with the 1 to 3 seconds typical of vision agents. My Grok Bot does all of its browser work through FSB now. I literally set up this Lounge profile by dictating to it over voice while it signed in with GitHub and filled everything out in real time, and it felt amazing. It worked great back when I was running my Claw too. Honestly, after living with it, the screenshot-based options feel a generation behind. It works on Chrome, Edge, Brave, Arc, anything Chromium. If your agent needs a site with no API, try it and tell me where it breaks. 🌐 full-selfbrowsing.com ⭐ github.com/fullselfbrowsing/FSB
Agents

0 comments

Log in to join the thread.