Claude Code

Scope: I tested the Claude Code CLI. I haven't used Anthropic's desktop or web app, so the notes below are CLI-specific.

I use it regularly for my day job. I run it in multiple terminals, use it for code reviews, infrastructure work, and research investigation. I have a suite of subagents, hooks, skills, and MCP servers I use with it. The daily ones are Linear, GitHub, Chrome DevTools, and GCloud.

One of my favorite things is using it to drive the browser to test and verify issues. When a bug report is hand-wavy, I have it reproduce the issue in a real browser and tell me what it found. That's faster than re-reading the report.

I also use it regularly to address PR feedback and fix issues others bring up. For most projects I start with plan mode to get a plan for how to implement and break things up. Then I switch to auto mode so it doesn't ask permission on every routine tool call.

The biggest thing I leverage is having multiple sessions running in parallel to work on different things at once, so I'm not waiting on it to respond. That's been the biggest productivity unlock for me.

It's expensive. I'm on the $200 Max plan and can still run out of session limits. I've gotten close to the weekly limit more than once. The session limit change in early May 2026 helped, and I haven't run out of a session since, but I still worry about the weekly limit some weeks. If you're not willing to commit at the $100 or $200 tier, the experience is a lot worse.

The ability to customize it and fit it to my flow makes it the best tool I've used so far for improving my productivity.