How I Use Claude Code Every Day
I’ve written about AI tools a few times on this blog, but I haven’t really talked about the one that has most fundamentally changed how I work. Claude Code ↗ is Anthropic’s CLI tool that gives Claude direct access to your terminal, filesystem, and development environment. At this point it’s involved in essentially every project I touch, whether that’s a web app, a native macOS app, Python scripts, workflow automations, or infrastructure configs.
CLAUDE.md#
CLAUDE.md is a markdown file you put in the root of a project that gives Claude Code context about your codebase — architecture, commands, conventions, gotchas. Claude reads it automatically when you open a project.
This is the single most important feature. It’s what turns Claude Code from a generic AI assistant into something that actually understands your project. It knows your build commands, your deployment pipeline, your file structure, and the decisions you’ve made along the way. Without it, you’re re-explaining basics every session. With it, Claude starts every conversation already up to speed.
Memory#
Claude Code also maintains a memory directory where it stores things it learns across sessions. API quirks, patterns that work well in a codebase, bugs that were painful to track down. The kinds of things that are easy to forget between sessions. This means Claude gets better at working in each project over time, and you don’t repeat the same mistakes.
Skills#
Skills are predefined workflows triggered with slash commands. The ones I use most:
/commit— Analyzes staged changes, generates a good commit message, and commits./commit-push-prgoes further and handles the full commit, push, and PR creation flow. I actually get better commit history from this than when I was writing messages manually./code-review— Reviews a PR for bugs, security issues, and whether changes follow the patterns in your project. Because it has the CLAUDE.md context, it understands your conventions rather than just applying generic rules./feature-dev— A guided workflow for building features. It reads the existing code, identifies patterns, and proposes an implementation that fits naturally into what’s already there. Useful for larger features where you want to think through the approach first./frontend-design— Builds production-quality UI components with Tailwind and shadcn/ui. It’s opinionated about avoiding generic-looking AI output, which is a common problem.
Plugins#
Firecrawl#
Firecrawl ↗ gives Claude the ability to search the web, scrape pages, and do research without leaving the terminal. Need to check API docs, look up how a library handles an edge case, or research a topic? Firecrawl handles it. This blog post used it for research.
Sentry#
The Sentry ↗ integration lets Claude pull up recent issues, dig into errors, and trace them back to the code. /sentry looks for open issues and helps resolve them. The Seer feature lets you ask natural language questions about your Sentry environment. The part I find most useful is the PR code review integration — if Sentry flags something on a PR, Claude can analyze it and fix the root cause.
Why a Terminal Tool?#
I know a terminal-based AI tool is not for everyone. Most people reaching for AI help with code are pasting snippets into ChatGPT or the Claude app, and that works fine for one-off questions. But the difference with Claude Code is that it’s already inside your project. It can read every file, understand the full directory structure, reference the CLAUDE.md and memory files, run your build commands, execute tests, and commit code — all without you copy-pasting anything.
That context is the whole game. When I ask Claude Code to fix a bug, it doesn’t just see the snippet I pasted — it can trace through the actual imports, check the actual database schema, and run the actual tests to verify the fix. When I ask it to build a feature, it reads the existing patterns and matches them. That’s fundamentally different from explaining your codebase in a chat window and hoping the AI infers the rest.
The agentic workflow is the other big difference. Claude Code doesn’t just suggest code, it writes files, runs commands, creates commits, and opens PRs. It’s doing the work, not just advising on it. Once you get comfortable with that, going back to copy-pasting between a browser and your editor feels slow.
What Makes It Work#
The CLAUDE.md gives it project knowledge. Memory gives it learning across sessions. Skills automate repetitive workflows. Plugins connect it to external services. Each piece is useful on its own, but together they make something that feels like a real development environment rather than a chat window.
It still makes mistakes, sometimes confidently. But the baseline quality with good context is remarkably high, and the speed at which I move through projects is genuinely different than it was a year ago. If you’re not using Claude Code, it’s worth trying.
