In plain English
An AI model only knows what's in front of it right now. Context engineering is choosing what goes there: the right instructions, the right files, a short summary of what's happened so far, and nothing that distracts. Too little and it guesses. Too much and it gets lost.
Example
A support assistant gets the customer's plan, their last three tickets and the refund policy with each message, instead of the whole help center. Its answers get more accurate and cheaper to produce.
In short
What the AI sees. Picking the right information to give a model, not just the right words.
Why it matters
For AI agents working through long tasks, what the model sees matters more than how the request is worded. Good context engineering improves accuracy, cuts hallucinations and lowers token costs. It has become a core skill for teams building with AI.
Common confusion
Context engineering vs. prompt engineering. Prompt engineering is about wording one request well. Context engineering covers everything the model sees, including documents, tool results and memory, often put together automatically.