I built Codepit, a desktop app for Claude Code, to see every change as it lands
Codepit is a free desktop app for Claude Code on Windows and Mac (not the snippet tool with the same name). It's my own project, not made by or affiliated with Anthropic. I work on a PC and a MacBook, and my Claude Code chat stayed on whichever one I started it on. I wanted the same chat on both. I

Codepit is a free desktop app for Claude Code on Windows and Mac (not the snippet tool with the same name). It's my own project, not made by or affiliated with Anthropic.
I work on a PC and a MacBook, and my Claude Code chat stayed on whichever one I started it on. I wanted the same chat on both. I also wanted to see the project while Claude works: the files, the editor and every change as it lands, not a summary in a terminal. So I started building that.
I usually have several Claude Code chats open on the same repo: one on the app, one on the website, and so on. Most of the time that works. One day it didn't.
One chat was putting out a release while another pushed a website change to master. The tag went onto the website commit, so v0.12.9 isn't a real release. I left it there, and the real one became v0.12.10.
Two days later I froze master. Now one chat merges and releases, every other chat works on its own branch and says when it's ready, and a pre-push hook refuses pushes to master and version tags from anywhere else. If you run several agents on one repo, I'd set that up on day one.
Keep your chats longer. Claude Code keeps chat transcripts for 30 days by default, then deletes them. To keep them longer, raise cleanupPeriodDays in ~/.claude/settings.json:
{ "cleanupPeriodDays": 3650 } Enter fullscreen mode Exit fullscreen mode If the file already has settings in it, add just that line.
Find add-ons with a real author. The Toolbox on the Codepit site lists Claude Code skills, mods and MCP servers. Each one traces to a real author and license and links to its source, so you can read it before you install it. You don't need Codepit for that.
It runs the Claude Code you already have, on your own Claude plan, so there's no second AI subscription, and the app itself is free. It's in private testing now. You can click through the demo and join the waiting list at codepit.app.
How do you keep track of what Claude changes while it works? I'd like to hear what you do now, and what's still annoying about it.