Claude Code review

What it is

Claude Code is Anthropic's coding agent. It runs in your terminal (with editor extensions available), reads your repository as needed, and carries out engineering tasks end to end: plan, edit, run, test, fix, commit.

Where it shines

Long autonomous tasks. Give it a genuinely large job — migrate a test suite to a new framework, thread a new parameter through forty call sites, port a module to TypeScript — and it works through it methodically, running the tests after each step and correcting its own mistakes. This is the class of work where in-editor assistants stall.

Being terminal-native turns out to be an advantage rather than a limitation. It composes naturally with git, build systems, linters and scripts, and it can be invoked from CI or a Makefile, which makes it useful for automation rather than only interactive help. It also reads files on demand rather than depending on a pre-built index, so it never works from a stale picture of the code.

Cost is favourable if you already pay for Claude: it is included in Pro and Max rather than being a separate subscription.

Where it falls short

The workflow is unfamiliar. Developers who live in a GUI find the terminal loop alien at first, and the editor extensions, while good, are less integrated than a purpose-built AI editor like Cursor.

Usage limits on the $20 Pro plan are the common complaint — a couple of hours of agentic work can exhaust them, which is why serious users move to Max at $100 or $200.

And autonomy cuts both ways. An agent confidently making changes across twenty files needs branches, code review and green tests. Teams without that discipline will get burned.

Pricing in plain English

Included with Claude Pro at $20. Max at $100 or $200 for 5x and 20x limits. API billing per token for automation and CI.

Who should use it

Experienced developers comfortable in a terminal who want to delegate whole tasks rather than autocomplete lines. Pair with Copilot if you also want inline completions while typing.