Point at anything on a live page. Handoff captures the screenshot, selector, styles and console — and hands it to Claude Code, Cursor, Codex or any MCP client. Or ask the agent to audit the page for you.
Click an element, drag a region, or record your screen — right on the live page. No tickets, no setup.
Screenshot, selector, computed styles, HTML structure, console errors, URL and viewport — grabbed automatically with your comment.
One command connects Claude Code, Cursor or Codex. The agent reads the full bug context over MCP — and fixes it in your codebase.
Ask the agent to audit the page: it hunts real breakage, or walks the page as a first-time visitor, and pins what’s worth fixing — marked as AI-created. You accept or delete. More on audit mode ↓
Not a screenshot in a prompt. Handoff speaks MCP: your pin and its context travel straight to the agent — which fixes the code and sends “resolved” back down the same wire.
pin · region · element
agent-ready
Claude Code · Cursor · Codex
Your agent gets what a developer would inspect in DevTools. Then it writes back: fixes the issue, notes the commit, marks it resolved. You accept or reopen.
Claude Code · via MCP: “Raised the button contrast to 4.6:1 (AA pass) in Button.tsx. Commit a3f9c2.”
You pin bugs for the agent to fix — and the same MCP connection works in reverse. Point Claude Code, Cursor or Codex at a live URL: it reviews the page like a careful reviewer, not a linter, and pins what it finds.
Objective breakage across three viewports — the things that are simply broken.
The agent walks the page as “a first-time visitor on a phone” and flags what would block that person.
Never lint noise. Capped to the few worth fixing, deduped against pins your team already left, every finding ships a real screenshot + selector. Nothing posts behind your back — you accept or delete.
Every comment lands on a shared Task Board — grouped by severity, agent-ready, with roles, invites and read-only share links.
You already ship with Claude Code or Cursor. Handoff feeds it the selector, computed styles and console errors automatically — prompts get shorter, fixes land faster, and the agent marks them resolved.
Handoff for developers →Review the live site, not a copy. Pin the exact element with its real rendered state — spacing, alignment, contrast — and every note carries the selector and a screenshot. No tickets written from memory.
Handoff for designers →Send a read-only link and anyone can follow the review — nothing installed, no account. Invite the people who leave notes and they get a free account and the extension. Everything lands on one Task Board, prioritized by severity and connected to real page context your team or your agent can act on.
Handoff for teams →A browser extension has to earn trust. Handoff captures only when you click — and the extension itself runs no analytics at all.
Yes — unlimited reviews and unlimited seats during the public beta. No credit card.
Any MCP client: Claude Code, Cursor, Codex and more.
Yes — ask it to audit a page. It hunts real breakage or walks the page as a first-time visitor, and pins only what’s worth fixing, marked as AI-created. You accept or delete.
No — they collaborate from the web Task Board, and anyone can view via a read-only share link.
No analytics, no data sales. Read the Privacy Policy.
Free during the public beta — unlimited reviews and seats, no credit card.