Claude Browser Agent

Claude tells you where to click.
What if it just clicked?

A Chrome extension that lets Claude act in your own Chrome: it clicks, types, uploads and checks its work, on the sites you're already signed in to.

Stop showing Claude what it already knows. No more screenshots, copy-paste or “here's what changed”. Claude looks at the page itself.

▶ See it work, recorded live: How it works (6 min) · A real game-dev day (6 min)

Would you use it? Get early access, and the early-bird price.

We're checking demand before we launch. Everyone on this list gets it for $29 once (early-bird). At launch it's $49. No subscription either way. We'll email you how it works on day one.

Email Name (optional) How often do you copy Claude's steps into your browser?

Every day A few times a week Rarely

What would you hand it first? Early-bird price for this list: $29 once (it will be $49 at launch). Would you get it?

Yes Maybe No

We only use your email to tell you about this product. No spam, unsubscribe anytime. Privacy.

You save more than tokens

Less back-and-forth. Building a site with Claude usually means: check localhost, take a screenshot, paste it back, explain what changed, check again. Claude now checks the page itself.
Less copy-paste. Errors, text, form values: Claude reads them straight from the page.
Less switching tabs. Claude does the clicking in Chrome while you keep working.
No tech skills needed. If you can ask Claude, you can use it. No code, DOM or APIs to understand.

In the founder's own Chrome over 12 days: Claude looked at the page itself 805 times (≈ 6.7 hours of screenshot-and-paste avoided, at an estimated 30 s each) and saved about 239M tokens (≈ $717 of AI cost), measured against sending raw pages and screenshots. Yours will vary.

How it works

You ask Claude (Claude Desktop or Claude Code), like you do today.
Claude acts in your Chrome through the extension, with its own on-screen pointer you can watch. Your mouse is never taken over.
Every action is verified. After each click or keystroke it compares the page before and after, and reports SUCCESS, FAILED or UNCERTAIN with evidence.
It stops for you at CAPTCHAs, one-time codes and payments, then picks up where it left off.
Reads pages as structure, not screenshots, so each step costs far fewer tokens.
You stay in control: one-click STOP, allowed sites only, and it asks before sends and purchases.

How is it different from Muse or Gemini in Chrome?

	Claude Browser Agent	Meta Muse	Gemini in Chrome
Where it browses	Your own Chrome	Its own browser, inside Muse ¹	Your Chrome
Your logins	Stay in your Chrome	Stored in Muse's credentials store ¹	Your Chrome session
The AI	Claude, the one you already code with	Meta's model	Gemini
Works with Claude Code	Yes: build the app, then set up the store in the same chat	No	No
Price	$29 early-bird · $49 at launch, one-time	See Meta	Auto browse is in Google's paid AI plans ²

1. Meta Help Centre, "How your Muse agent browses the web". 2. TechCrunch, 28 Jan 2026. Details as publicly described at the time of writing; check each product for current terms.

Claude tells you where to click.What if it just clicked?

You save more than tokens

How it works

How is it different from Muse or Gemini in Chrome?

Claude tells you where to click.
What if it just clicked?