I built a tool to verify what your coding agent says it fixed

Hi fellow devs! There's a gap that kept biting me: the agent says "done", and then I have to figure out what to actually check. On one project I ended up with 72 things to test and nowhere to keep track of the results. That's how PalmtopAI was born: a checklist shared between you and your agent. The agent writes the checks to do, you test them from your phone (or any browser) and mark what works, what doesn't and what leaves you in doubt. Then the agent reads your review and fixes what's broken. It works with Claude Code and Codex (and other agents that run shell commands), and needs macOS, Linux or WSL and Node.js 22+. Checklists stay on your computer, the relay only forwards messages. It's an early preview and I'm building it solo, so I'm looking for people to try it on one of their projects and tell me where it gets stuck: installation, phone pairing, anything. If you get stuck, just message me! Suggestions and ideas are very welcome too: what's missing, what you'd change, what you'd need to actually use it. 👉 https://palmtop.mtwa.it/