The word trips people up because they picture “the code” as one giant document. It usually isn’t. A codebase is a collection: pages, styles, scripts, configuration, data, tests, docs, and tiny helper files that only make sense together.
We often explain GitHub as Google Drive for code, and the same picture works here. A shared drive folder might hold documents, spreadsheets, images, instructions, and old versions. A codebase is the software version of that folder: the files an app or site needs to run.
How it shows up
When you open a repository, you’re usually looking at a codebase. The repository is the managed container; the codebase is the actual project inside it, and git tracks how it changes over time. When you use Claude Code, Codex, or another coding agent, the codebase is what the agent reads. It may hunt for the file that defines a button, inspect a failing test, or add a new page. This is why agents explore before editing: a good agent reads the existing project for naming patterns, helpers, tests, and conventions instead of guessing from memory. A messy codebase gives it less structure to follow, the same way a disorganized shared drive slows you down.
Why you care
You don’t need to become a developer. You just need to know that a software project is a folder of connected files, and changing one file may affect others. AI can only change a project well if it can understand the project as a whole, and the codebase is that whole.