AI Glossary
Coding agent
coding agent, programming agent, code agent, AI coding agent, software agent
A coding agent is an AI agent that writes, runs and fixes code on its own, working in a loop with a developer's tools — it reads files, runs commands, reads errors and applies fixes until it reaches the goal.
- It doesn't just suggest code — it runs and fixes it in a loop.
- Uses developer tools: a file editor, a terminal, tests, version control.
- Claude Code is one example, working in the terminal and the project repository.
A coding agent is a specialized AI agent whose job is to work on code. It is given a task described in plain language, then plans the changes itself, reads the existing files, applies fixes, runs the code or tests, and decides what to do next based on the result. This ability comes from tool use: without access to an editor, a terminal and version control, it would be nothing more than a generator of code snippets.
The difference from ordinary code completion is significant. A copilot AI built into the editor suggests the next lines as the developer types — the person is still driving the work. A coding agent runs the task through to completion on its own and verifies the result itself, which makes it a realization of the broader paradigm of agentic AI. Claude Code is one example, working in the terminal and the project repository.
In day-to-day company work, a coding agent takes over repetitive, well-defined tasks: migrations, fixes based on a test report, refactors, or implementing a feature from a detailed specification — often as part of a broader agentic workflow. The safety boundary remains a human review of the changes before they are merged, along with control over which commands the agent is allowed to run.
Related terms
Related articles