LLMs are already capable of writing code. But generating a single code snippet is very different from building real software. Real development involves understanding requirements, navigating large codebases, writing and modifying multiple files, running tests, debugging errors, and iterating until the solution works.

Coding agents are designed to automate this process.

A coding agent is an AI system that can reason about programming tasks, interact with a development environment, and iteratively improve code. Instead of producing one-off answers, the agent can read files, plan changes, write code, execute it, analyze failures, and refine the implementation. This allows AI systems to assist with complex engineering workflows such as implementing features, fixing bugs, refactoring code, and generating tests.

In this chapter, we will explore how coding agents work, the tools they use to interact with codebases, and the architectures that enable iterative development loops.