A coding assistant is an AI tool that helps you program while you work. It can complete the line you are typing, explain an error in a chat, apply a change you ask for or, in agent mode, chain steps on its own. Well known examples are GitHub Copilot, Cursor and Claude Code.
✅ Tests pass
A question from the app. Pick an answer.
What do a coding agent's tools enable?
An agent makes mistakes too. The difference is that it can test what it did and see the error.
It works on the real files: it does not tell you what to paste, it changes it and can check it.
It can use the same kind of model as a chat. What changes is the tools it has at hand.
One tool usually combines several ways of working. Suggest: it completes the line and you accept it or not. Edit: you ask for a change and it applies it to the file. Act: as a coding agent, it reads the project, edits, runs the tests and checks the result, several steps in a row. The more it does on its own, the more reviewing the diff matters.
Inside there is a language model: it can suggest code that looks right and is not, or make up a function that does not exist. Tests and your own reading are the safety net.
When a job ad asks for "daily use of coding assistants", it wants you to program with one without losing control: give it clear context, ask for small changes, read what it suggests and test it before you accept it. To show it, bring an example: a project where you used one and something you had to fix.