The clearest expression of the agentic approach: no editor to switch to, just a command in your terminal that can read the project, write files, run the build, read the failure and fix it. Because it works through the same tools you do, git and your test runner included, it handles the jobs that span a codebase rather than a file, such as a migration, a refactor with a passing test suite at the end, or tracing a bug you cannot reproduce by reading.
It suits people comfortable delegating: you describe the outcome, review a diff, and iterate, rather than watching each keystroke. That means it is less useful as a typing accelerator than an editor-based tool, and it works best on projects with tests and a tidy structure it can verify against. Pair it with your usual editor rather than replacing it.