CLI agent
Claude Code
Anthropic
Claude Code runs in the terminal. You point it at a project, describe a task, and it can read files, propose edits, and run shell commands. Access is through a paid Claude plan or API usage. Confirm what your plan includes on anthropic.com.
8.4
Preliminary
Editorial score out of 10. Not a benchmark.
What we tried
Ran it in a terminal on a small repo and asked it to adjust a component and explain the diff. It stayed in the project. First impression only.
Scores are preliminary editorial notes from a short hands-on pass. They are not benchmark results, not pass rates, and not for sale. A shared task suite is not published yet, so we do not print speed, cost, or success-rate figures.
Preliminary notes
- Setup7/10
- Editing8/10
- Agent work9/10
- Docs and control8/10
Good fit
- Repo tasks you are happy to drive from a terminal
- Longer changes where a written plan helps
Limits
- You need a paid Claude plan or API credits
- It will run commands if you let it, so the permission prompts matter