I started out designing websites, moved into product design, and ended up founding Huddlekit. It's the tool I kept wishing I had.
For years I sat right between design and development. I'd design a site, hand it over, and then spend the rest of the project chasing what happened to it on the way to launch.
Feedback came in as screenshots in Slack, notes in email and comments like "the button looks off". Nobody knew which page, which breakpoint or which browser. Every project ended with a Friday scramble to work out what had actually been agreed.
Nothing was built for the handover
The tools we tried were made for one side of the table. Designers got somewhere to leave comments, developers got a list they couldn't act on, and clients got a browser extension they didn't want to install. Most of them charged per seat for the clients too.
So we set out to build a tool shaped around how designers and developers actually work together, while the site is being built and after it's handed over.
Paste a URL, send one link, and anyone can comment right on the live page.
That's still the core of Huddlekit. Every comment lands with its screenshot, breakpoint, browser and selector, so a developer can fix it without a follow-up. Clients don't need an account, and guests never count as seats.
Agentic first
The way we all build has changed. More and more of the fixing happens with an AI agent next to the code, and we believe a review tool should be built for that at its core, not bolted on at the edges.
That's why Huddlekit speaks MCP. Claude, Claude Code or Cursor can read a comment with its page, element and screenshot, and make the fix. Give it write access and it replies on the thread and marks the comment resolved. Your team decides what ships, the agent does the legwork.
We ship improvements every week, and most of them start with someone telling us what got in their way. If something slows you down, we want to hear about it.
