Cursor
The AI-first code editor that made agentic coding mainstream.
Coding
Anthropic's terminal agent: describe the change, it reads the repo and edits the files itself.
Claude Code runs in your terminal, reads the repository, plans the change, edits files, runs tests and iterates until the tests pass. It works on real codebases rather than snippets, and it asks before doing anything destructive.
It is strongest on refactors, test backfills and well-scoped bug fixes inside a codebase that already has tests. Give it an ambiguous product requirement with no tests and it will produce a confident change you cannot verify.
Main job: Delegate a scoped engineering task to an agent that can edit and verify it.
Refactors, test backfills and bug fixes in repos that already have tests.
Point it at one directory, ask for a plan first, then approve step by step. Keep the repo clean and commit before you start so a bad run is one `git reset` away, and let it run the test suite itself rather than describing what it would do.
Watch out: it can run shell commands, so treat it as a junior engineer with your credentials. Never run it against production config or a repo with live secrets, and read the diff before you merge — passing tests are not the same as correct behaviour.