Coding agents leverage three core components: tools, memory, and repo context. By calling external APIs, they extend LLMs beyond text generation into real‑world actions. The memory module stores past interactions, letting agents remember user preferences across sessions. Repo context feeds file snippets, enabling accurate edits and code synthesis, reducing debugging time.