Let your agent see what it built.

So it can visually understand, verify, and react by itself, to achieve your goals.

Browsershot captures pages and browser state from the command line.

Capture a page

$ browsershot https://example.com --inspect body --with-errors --json
# {"outputPath":"...","consoleErrors":[],"inspected":{...},...}

Check the page as it runs

Capture a page after an action and inspect an element in the same run. Your agent can use the results to decide what to fix next.

For example, open a menu and inspect it

$ browsershot https://app.example \
    --act 'click:#menu' --inspect '#menu' --json

ScreenshotSee the page after the action.

JSONRead the inspected element and capture details.

Pull requests: Capture screenshots for reviewers without repeating the work by hand.
Repeatable agent work: Give your agent a capture command instead of writing browser scripts or driving a desktop browser. It can save time and tokens.
Evidence and records: Capture screenshots for compliance reviews, audits, or other records.

Use what fits the task. Each option works with a capture.

Pages and elements: Capture a whole page or a selected element. Wait for text or an element before capturing.
Browser errors and requests: Collect console errors, page exceptions or network activity when you need to investigate.
Component libraries: Capture named Storybook and Ladle examples.
Saved pages: Reuse routes, settings and named elements.

Capture pages behind a login using a named user.

After setting up Authstate

$ browsershot https://app.example \
    --auth-user member

Install from source with Bun and Git, a Chromium download and a local build. The GitHub guide has the steps.