Hey there, shippers!

I’m back, and bringing some fresh ideas with me. New experiments, new tools, and new ways to ship smarter (and faster). Let’s go!

I’ve been playing with Claude Code on the Web using two personal projects, one in iOS and one backend, and the results were quite satisfying and fast.

The setup is simple: you add Claude to the repository and it’s ready to analyze your code. It uses the same integration as when you enable Claude to review PRs.

In my tests, I wrote a short paragraph describing the changes I wanted, and it handled them well in both repositories without needing extra instructions.

In the TypeScript project, it even ran npm commands inside the sandboxed environment.

In the iOS project, it can’t compile or run the app on its own because it runs on Linux, which was expected. My workaround was to have it open a PR and then analyze the build and test results from that PR, which run on my self-hosted Mac mini. It’s something!

Something I’d really love to see is support for workflow configurations, so it could run directly on self-hosted machines or in custom environments.

Two really cool features: it can create a PR with a single click, and you can resume the same session directly from the CLI (using claude --teleport <session_id>), which is very convenient. I’d say the web experience is practically the same as the CLI, minus the speed.

It’s still in beta and not available for all plans, and the privacy/data retention question needs to be verified. But at the pace these AI tools are being released, this should reach everyone soon. It’s a great addition to the toolset we can use, especially in those moments when you’re away from home, without your laptop, but still want to keep shipping.

One thing that puts me off is their interface in general… I really don’t like it.