What is a context window, and why your memory files fit inside it

The context window is the total amount of text Claude can actually see when it generates a response — your message, the conversation history, any files or memory files you’ve given it. If something isn’t in the context window, it doesn’t exist to the model for that response, no matter how obviously important it is.

This is the practical reason memory files work at all: PRODUCT.md, USERS.md, GOALS.md, and VOICE.md get read into the context window every time you work in that project, so Claude doesn’t need you to re-explain your business — it’s just there. It’s also why a founder who describes their situation in one rushed sentence gets thin output: that sentence is the entire context window for that answer, and the model can only work with what it’s given.

Context windows are large, but not infinite — check current capabilities in Claude for exact limits, since these expand over time. For most founder use, the practical lesson isn’t “don’t run out of space,” it’s “put the real thing in, not a summary of it.” A messy but complete spreadsheet in the context window beats a tidy paragraph describing the spreadsheet, every time.