Claude Code tips from the creator of Claude Code
Claude Code is crazy customizable, and its creator just shared his setup
The best thing about Claude Code is that it works incredibly well out of the box.
The second-best thing is that it is quite customizable.
You’ve probably seen our guides on Claude Skills, Hooks, and other customizable things already. But today’s article is uniquely valuable.
Boris Cherny, the creator of Claude Code, recently shared how he set up Claude Code. I’ve parsed the entire list, and compiled the highest-value tricks in this list, plus some additional context.
Use /permissions for an allowlist
It took me too long to learn this. Claude Code works on permission questions by default. If you want to get the most benefits out of agents, you can’t be constantly watching the agent do its work.
That makes it impossible to run parallel agents (more on that later).
Claude has a skip permissions flag, but that’s really the nuclear option, especially on a work computer.
Instead, you can make an allowlist of commands that you know are safe in your environment, to avoid constantly getting hit with permissions prompts.
Start by running /permissions - then you’ll see an interactive prompt:
Here, you can add to or manage existing commands in your allowlist and denylist.
These will go in .claude/settings.json
