Memory & context
What is Context?
Context is the information placed in the model call for this step.
It may include the user task, retrieved memory, tool results, and instructions. Context engineering is choosing that set.
Why it matters
Context is everything the model sees on a given call: the system prompt, the conversation history, retrieved documents, tool results, and instructions. The context window is finite, and what you put in it determines the quality of the output.
Context engineering — choosing what goes in — is one of the most impactful skills in agent development. More information is not always better; relevant information is.
Key takeaways
- 1Context = what the model sees for this call. It directly determines output quality.
- 2Context windows are finite — relevance beats volume.
- 3Context engineering is choosing the right information for each step.
Common mistakes
- ✕Stuffing the context with everything available instead of selecting relevant information.
- ✕Not summarizing long conversations before they overflow the window.
Related terms
Concept neighborhood
Terms linked from Context in the glossary graph.
- Context
- Context Engineering
- Memory