index

Claude Code: One Month Review

· 3min

My review of Claude Code after a month of active use.

I barely finished this post when Google released their competitor Gemini CLI, but we’re not rushing — there’ll be a post about that too if I test it myself. For now, Perplexity analyzing the docs and Reddit gave this comparison (says it’s still raw).

For those who already know everything below: did you know you can use Claude Code with any model, even Gemini?? Now you do)

Let’s continue. So, a month ago I added Claude Code to my toolkit and honestly — didn’t expect such a difference. Cursor stayed as my editor, but some tasks that used to take hours now get done in 20-30 minutes.

How is it Different from Cursor?

The difference is in the architectural approach: Cursor is an AI-integrated code editor with a GUI that works with local context and is optimized for token savings and quick responses within the $20 plan. Claude Code is a terminal agent (CLI) that autonomously explores the entire repository, understands the project’s overall architecture, and executes multi-step tasks without saving resources — using full context up to 200K tokens, which allows it to generate deeper and more architecturally sound solutions.

Plus Subagents for big tasks. If the task is complex, Claude can spin up multiple sub-agents in parallel. Instead of an hour of work — 10-15 minutes.

Most Useful Features from Practice

Plan Mode (shift+tab) — essential. Plan first, then code. For complex tasks I add “think hard” or even “ultrathink” — the agent spends more time on analysis.

Claude.md file in the project root — this is the agent’s memory. I record project rules, architectural decisions, code style; it can supplement it and create other .md files as documentation (it loves them).

Custom commands in the .claude/commands folder — I create markdown files with frequently used prompts. For example, /git-commit which generates commit messages following our standards, or /update-doc that scans changes and patches documentation. Examples of slash commands here.

MCP integrations are available. Same as in Cursor, but like Claude Desktop itself, Claude Code somehow uses them better.

My Workflow

Explore → Plan → Code → Commit. Frequent commits replace the missing restore function from Cursor.

And yes, it’s CLI, but they’ve done great work on UX — it doesn’t feel like a regular terminal. It’s very interactive — you can drag screenshots directly into the terminal. For those who need more, there’s Claudia. It also integrates nicely with VS Code/Cursor, and I generally use them together and switch often. Claude Code works great in the background and has better “overview,” while Cursor is more interactive.

The Downside

The main minus is price, but they recently included it in the regular 20subscription,andtheMaxsubscriptionat20 subscription, and the Max subscription at 200 gives 20x more tokens (hard to hit the limit). Or you just pay per token.

Useful Resources