FRENES

Glossary · Agents and autonomy

Coding agent

A coding agent is software that combines an AI model with tools to perform programming tasks. From a request you provide, it can inspect a project, change files and run checks, subject to the capabilities, access permissions and execution limits of its working environment.

Coding agent: the explanation

You describe a problem or a change you want. The agent locates relevant files, reads project instructions and decides how to approach the task. It can then edit code and use a terminal to run permitted commands. Error messages and verification results inform its next actions. Depending on the setup, it may work in your existing folder, a separate copy of the project or a remote environment created for the work you have requested.

The tool connects an everyday description of a task with several development operations. It might investigate a defect, adjust an interface or explain how related files work together. Its ability to act extends beyond returning a sample program in a conversation. However, the label covers products with different scopes. Some only prepare changes for review, while others can also interact with external services or take part in a release process when the necessary permissions exist.

Generated code can look reasonable while containing a defect or adding an unnecessary dependency. Checks should therefore address the behavior you need, and passing them only provides evidence for the situations examined. The agent also needs access that matches the assignment. Reading a project does not always require its live credentials. You should be able to inspect the changes, understand the checks performed and distinguish editing a local copy from deploying a running service.

A concrete example

Your club’s registration form rejects some names containing a hyphen. You ask the agent to investigate and prepare a correction. It finds an overly restrictive validation rule, updates the relevant file and checks several representative entries. It then presents the change and explains what it verified. You or a reviewer can assess the result before it reaches the website used by club members.

Why it matters

A coding agent may change components that your service depends on. When choosing one, examine its permissions, the clarity of its changes and its approach to verification. A successful demonstration on a small example does not establish reliability across every project. Access controls, protection of confidential information and a practical way to reverse a change remain central to using it responsibly in real work.

Common misconceptions

  • A program that runs without an error is not necessarily correct, because it may calculate the wrong result, overlook an important case or introduce a defect that the checks performed do not reveal.
  • A coding agent does not automatically need all your account permissions, since many changes can be prepared and reviewed in a project copy without deployment access, production credentials or authority to publish the result.

Related terms: agents and autonomy

AI agent
An AI agent is a system that uses a model to decide which steps to take toward a task.
autonomous agent
An autonomous agent is an AI agent that can pursue a goal and choose several actions without involving you at every step.
multi-agent system
A multi-agent system is a collection of agents that divide tasks or exchange information to pursue an objective.
planning (AI agents)
Planning in AI agents is the preparation of steps an agent intends to follow toward a goal.
agent memory
Agent memory is the set of mechanisms that lets an AI agent retain and retrieve information useful to its work.
human in the loop
Human in the loop is an arrangement in which a person participates at selected stages of an AI system’s work.
tool calling
Tool calling is a mechanism through which a model asks software to run a specific function, such as searching documents or checking a calendar.
computer use
Computer use is a capability that lets an AI agent interact with a graphical interface by clicking, typing or scrolling.

Where this site covers it

The term appears on 294 pages of this site. A few of them:

Same family

Go further

Frequently asked questions

Do you need programming knowledge?

You can describe a task without writing code. For consequential changes, someone able to review the result remains useful, particularly when checking expected behavior, access to information and the effects on the wider service.

Can it change several files?

Yes, when its tools and permissions allow it. One fix may affect code, configuration and documentation. You should be able to inspect the complete set of changes to judge their consistency and overall scope.

Does it publish changes automatically?

That depends on the setup. Preparing edits, recording them in version control and deploying them are separate operations. Check which actions are authorized and which need your confirmation before the agent starts its work.

Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score