Strip away the jargon and it’s a folder with better memory. The files, the instructions, the history, and the current changes all sit in one place.
Think of it like Google Drive for code. In Drive you might have a client folder with proposals, spreadsheets, notes, and a README-style note telling a new person what they’re looking at. A repository is that same idea, built for software and agent work. It holds the codebase, but it also remembers the changes people made, who made them, and how they connect to the project.
A repository is the place where a project can be worked on without losing the thread of what changed.
How it shows up
That history is the piece Drive doesn’t handle cleanly. Ask an agent to edit a file and you can see exactly what changed, then review it, fix it, or throw it away. If someone else is working too, a branch keeps their work from colliding with the main version. When you use Claude Code or Codex, the repository is the workspace. An agent in the wrong one is like a new hire sent to the wrong shared drive: smart, but reading the wrong files and answering confidently from the wrong context. GitHub is the hosted place where many repositories live, the way Drive isn’t the folder itself; it’s where the folder gets stored, shared, and reviewed, and a README is the first note you look for.
Why you care
For client work, the repository creates accountability: you can ask “what did the agent change?” and get a real answer. Agents are only useful when they’re standing in the right room.