My bug reports and feature ideas were scattered across camera roll screenshots, a physical notebook, Notion pages, Google Keep, starred emails, WhatsApp messages to myself. They would keep piling up and rarely get acted upon.
I wanted a simple feedback widget to unify it all—give myself and my users a proper way to report stuff so nothing gets lost.
The existing tools sucked. Usersnap is 518KB. Hotjar is 230KB. They're bloated, ugly, and look "bolted on." The dashboards are a mess. Not something I'd want on a polished product.
So I built my own. Usero.io is 7.6KB, 30x smaller than Hotjar, (in my opinion) actually looks good, and has a modern dashboard.
Then I went 1 step further. I wired up Claude to analyze my codebase and write fixes. Feedback comes in → PR gets opened automatically. So far I've merged the PRs as-is 85% of the time, but even when it's wrong, it's a useful starting point.
Here's what I realized: The bottleneck was never collection. It's action. Between "user reports issue" and "code gets written" there's just... a human sitting there context switching.
That "10 minute fix"? It actually eats 45+ minutes across two people. Reproduce it, find the file, fix one line, test it, open a PR, get it reviewed, merged, deployed. Plus you lose the thread on what you were actually working
on.
Most tickets don't need judgment. They just need execution. If the task is clear, the human can skip whole loop and come in at review time.
What DOES need a human: cross-team design, large architectural decisions, security issues, matters of "taste" (what should the design aesthetic be?). Those require your brain. But probably >50% of dev work doesn't.
Usero is built on Cloudflare Workers + D1 with Claude Code running in an isolated container for each PR (so your repo isn't stored or leaked). I'm working solo in Vancouver, bootstrapping this thing.
Pricing: Free tier gets you 5 PRs/month and 1,000 feedback items,. Pro is $19/month, way better value than the enterprise solutions. It's cheap to store your feedback.. why should you pay an arm and leg? The widget itself is MIT open source.
Would love feedback from folks here—what's missing? What would make you give it a go?