Definition
Context Engineering is the systematic software practice of sourcing, filtering, structuring, and sequencing dynamic information payloads supplied to an LLM's context window for a given task.
Purpose
Solves the "needle in a haystack" problem and context window bloat, maximizing model reasoning fidelity while minimizing inference latency and token costs.
Key Concepts
How It Works
Systems pull data from databases, active states, and RAG pipelines, compressing and ranking the payload into structured formats before dispatching the model prompt.
Related Terms
Sources
- Birgitta Böckeler, AI Agent Harness, Thoughtworks
- DomainDrivenUX Context Architecture