What is Context Engineering?

Context Engineering is the design of everything an AI model or agent "sees" at the moment of a decision: instructions, data, tools, memory, examples, and rules. Rather than only optimizing the wording of a request, it defines which information is provided in what structure, order, and amount. The aim is for the model to receive exactly what is relevant – neither too little nor too much.

Why does this matter?

An agent’s quality depends less on the model than on the context it receives. For process questions this means an agent needs cases, objects, timestamps, metric definitions, and responsibilities in structured form – not a pile of documents. Provide that process context cleanly and you get dependable answers; leave it to chance and you get plausible ones.

What does this look like in practice?

An agent is asked to explain delivery delays. Without context engineering it gets the question and database access and has to guess tables, field names and metrics. With context engineering it receives the definition of "late" (date per order confirmation, not per purchase order), the object model of order, delivery and plant, and a tool that returns cycle times per variant — and answers in two steps instead of twelve.

Not to be confused with

Not to be confused with prompt engineering, which optimises the wording of a single request; context engineering shapes the whole information environment in which an agent works over many steps. Nor the same as RAG: retrieving text passages is only one of several techniques for supplying context.