Install it.
Just once.
Add the component to your Next.js layout, a validation route, and a server-side secret. You’re ready.
For the developerOne link to open your site. One pen to show what’s wrong. All the context your agent needs to figure it out.
Brushed aluminum. Bright ideas.
The issue, the screenshot, and everything behind it.
You show the problem.
Your agent gets the context.
You move forward together.
Add the component to your Next.js layout, a validation route, and a server-side secret. You’re ready.
For the developerAsk for a link to the page you want to test. Your agent adds the recipient, version, or ticket. You share it.
One link, one sessionOpen the link, circle something, type, or speak. Browse your site and collect your feedback in one report.
For everyoneSend the report. Allow browser control so your agent can inspect, click, and reproduce the bug to fix your code.
With your permissionSkip the “which page?”, “what screen size?”, and “can you open the console?” Your feedback comes with the clues that matter.
Give my agent the context ↗A page capture, drawings, notes, and voice recordings.
Console output, JavaScript errors, and network requests captured after activation.
URL, viewport size, DOM details, and metadata from the link.
Next.js App Router or Pages Router.
A secret that stays on the server.
A quiet tool, activated on demand.
Early release. Version 0.0.2 is now available on npm. View on npm ↗
The report is enough to get started. If your agent needs to dig deeper, allow read-only access or browser control. It can inspect the page, take screenshots, click, and fill allowed fields. You can turn access off.
Yes. Install the package and configure your secret on your Next.js server. Only a valid signed link activates the widget in the recipient’s tab. It does not replace your site’s authentication.
No. They open your link and the widget appears on your site. They can draw, type, and record a voice note after granting microphone permission.
Your agent keeps the report link when it creates the invitation. Once the user sends the report, the CLI downloads screenshots, notes, and logs. With permission and the tab open, your agent can also interact with the page live.
Any agent that can run the CLI can use it, including Claude Code and Codex. A chat without terminal access cannot control the browser through the CLI.
Annotated page captures, your notes, errors, and fetch/XHR activity after activation. No cookies, headers, or request bodies are collected. Captures use DOM rendering, so videos and some external content may be missing. Review your report before sharing.
A tiny tool for better work together.