AI coding assistant vs Coding agent

Why they differ

An assistant answers and suggests while the developer drives; a coding agent is handed a task and carries out many steps, runs commands and edits files by itself.

AI coding assistant

AI-assisted coding

A tool built into a programmer's editor that suggests, explains and rewrites code using a large language model.

Formal

Software that sends the code a developer is working on, plus nearby files and a request, as context to a large language model and shows the answer as inline code completion, chat replies or proposed edits that the developer accepts or rejects.

In plain English

Like a helper who has read a great deal looking over your shoulder while you write - quick with ideas and happy to explain, but it has not seen your whole project and is sometimes confidently wrong.

In practice

A newly hired developer in a region's IT department asks the assistant what an old function in the lab system does, gets a short explanation, then has it draft the missing input checks, which she reads and adjusts.

Why it matters

It speeds up routine work, but whatever code and secrets are open in the editor may be sent to an outside provider, and its suggestions can carry the same flaws as the public code it learned from.

Coding agent

AI-assisted coding

An AI that is given a programming task and carries it out itself - reading files, running commands and editing code until done.

Formal

An AI agent whose tools act on a code project - reading and writing files, searching, running shell commands and tests, and using version control - and which loops through plan, act and check steps via tool calling until the task is done or it hands back to a person.

In plain English

Like leaving a carpenter alone in your house with a list of repairs - he gets on with it by himself, but he can also open any cupboard you left unlocked.

In practice

A developer in a municipality asks the agent to add a date filter to the report page of the housing benefit system; it finds the files, makes the change, runs the tests, fixes one failure and opens a change request.

Why it matters

It can do hours of routine work alone, but it runs real commands with the developer's rights, so a mistake or a planted order can delete work, leak keys or push harmful code.

Shared connections

Code completion Vibe coding Secrets management