Claude Code is an agentic coding tool from Anthropic. You give it a task in plain language; it reads your codebase, edits files across it, runs commands and opens pull requests — on your machine, against your real project.
It is deliberate by temperament. It plans before it acts, explains what it is doing, and stops to ask when a change is bigger than the instruction implied. That is friction when you wanted the job done unsupervised, and exactly right when the job is “work out why this only breaks in production”.
It cannot see what it builds. It reads and writes files and has no picture of the page those files produce — which is fine for a migration and the central problem for anything with an interface.
What it is
Claude Code is not an autocomplete extension and not a chat window that hands you snippets. It is an agent: you describe an outcome, and it decides which files to open, what to change, which commands to run and when it has finished. The unit of work is a task, not a line.
That distinction matters most on work that spans a codebase. Asking for a function is something any assistant can do. Asking it to rename a concept across forty files, update the tests that assert on it, and leave the public API untouched is a different kind of request — and it is the kind Claude Code is built around.
Where it runs
The same engine runs across a lot of surfaces, and a session can move between them — so work started on a laptop can be picked up from a browser later.
| Claude Code | |
|---|---|
| Made by | Anthropic |
| Models | The Claude family — Opus, Sonnet, Haiku and Fable |
| Surfaces | Terminal, VS Code, JetBrains, a desktop app, the web and mobile |
| Access | A Claude subscription, or an Anthropic API key |
| Project instructions | CLAUDE.md, plus memory it writes itself as it learns your project |
| Extending it | Skills, hooks, MCP servers, subagents and an Agent SDK |
| Unattended work | Routines on Anthropic's infrastructure, background agents, GitHub and GitLab CI |
Telling it your rules
Every agentic tool needs somewhere to keep the things you would otherwise repeat in every prompt: the test command, the paths not to touch, the commit message style, the fact that this project uses tabs.
Claude Code reads CLAUDE.md from your project root at the start of every session. It also builds its own memory as it works — noting build commands and debugging findings without you writing them down. If you also run Codex, you maintain AGENTS.md alongside it; most people keep the substance in one file and have the other defer to it.
What it is good at
Worth being specific, because “it writes code” is true of everything and useful about nothing.
- changes that span many files, where the risk is a wrong assumption rather than slow typing
- debugging something whose cause is not obvious yet
- codebases with conventions that matter and need following, not guessing at
- work you want to watch, and interrupt at step two rather than undo at step ten
- wiring agents into your own tooling with hooks, skills or the SDK
- anything whose success is judged visually — it cannot see the rendered result
- spatial instructions, where prose is a lossy way to describe a change you can already picture
- fully unattended work, if you wanted no narration and no questions
- a task specified loosely enough that “more prominent” could mean five different edits
Seeing what it builds
The limitation in the right-hand column above is the one that has no prompt-level fix. Claude Code's only representation of your page is the source text; yours is the rendered picture. Every round trip pays that translation cost twice — once when you turn the picture into a description, and once when the agent turns the description back.
That is the specific gap Clodex exists to close: it runs Claude Code against your real project, renders the result in a live preview, and puts design tools on that preview, so you select the element you want changed and change it by hand. The edit is written back into your real source as ordinary code. The visual editor guide covers what that looks like in practice, including the parts it does not do.
Guides
See what it builds.
Run Claude Code on a live preview of your project, and edit the result by hand.
- Claude Code documentation — Anthropic's own overview of surfaces, memory, skills, hooks and subagents.
- Clodex security — where credentials live, what the agent can reach, and how to report a vulnerability.