Context is the information the AI can see right now while it works: the working material, distinct from memory in the human sense. In a Claude project, project knowledge is context you attach on purpose. In a vault, it’s notes, transcripts, rules, and prior decisions. In RAG, the system retrieves relevant chunks from a knowledge base and brings them into view.
The papers on your desk make this concrete. The contract, notes, and checklist in front of you let you work; the same files locked in a cabinet down the hall don’t help. Pile every company document on the desk and the important page disappears.
Why you care
Better context usually beats a cleverer prompt. Give the AI the wrong papers and it produces the wrong work with confidence; give it the right ones and the work gets sharper, faster, and easier to review. Good context is the right context for the job, sized to fit the context window. Organize the right files and an agent picks up the work fast. Leave them scattered and it guesses or burns time searching. Making the work visible matters far more than sounding technical.