ReplayAutonomous app testing for teams who ship at AI-speed
Promoted
Launching today
Browsentic
No new browser, no API key. Your tabs, your own AI agent
1 follower
No new browser, no API key. Your tabs, your own AI agent
1 follower
Like Claude Code for your browser, except it can see the page. Record a job once and it replays it, explores unfamiliar sites on its own, drives several tabs at once, and teaches itself skills you write. Nothing leaves your machine. Most of all it uses a loopback connection to your AI agent harness you already use (claude, codex or antigravity)
Hi Product Hunt,
This is Shahriar, author of browsentic.
Every AI browser wants you to switch browsers and log in to everything again, or hand over an API key and pay per token on top of the subscription you already have.
Browsentic does neither. It's an extension for the browser you already use, running on the agent you're already signed in to - Claude Code, Codex or Antigravity, switchable with one click. Think of it as Claude Code with eyes on your browser: same agent, except now it can see the page and act on it.
Show it once, and it remembers. Hit record, do the job yourself, and later just say "do it like last time." It replays the intent, not a brittle script - it re-checks every element on the live page as it goes.
It finds its own way around. Point it at a site and it explores, then writes itself notes. Every session after that already knows where the search box is and how the results load.
It doesn't stall on the boring blockers. When a "verify you're human" wall appears, it recognises the widget, ticks the box with your approval, and hands you anything that needs a real person - instead of silently failing like most automation does.
It works in parallel. Three tabs at once by default, up to eight - real multitasking, not a queue.
Point at what you mean. Press the lens and click the thing you're talking about. It works both ways: when words aren't enough, the agent hands the lens back and asks you to point.
Teach it your own moves. Write a skill in plain markdown, drop it in, and it becomes part of what the agent knows.
Security is built in, not bolted on. A fresh install contacts nothing until you redeem a one-time pairing code. Everything runs on your machine - no cloud, no account, no API key. A policy gates consequential actions, keeps each run on the sites it's about, and treats all page text as untrusted data, so a web page can't talk your agent into something. Tune any of it from Settings.
`npx browsentic setup` - MIT, source is all there.
Source: https://www.github.com/imshaikot...
Website: https://browsentic.com