Glossary
What is Context Engineering?
Context engineering is the practice of giving AI agents the right context at the right time. Coined by LangChain and adopted by Anthropic, it covers context window management, retrieval design, prompt structure, and the systems that supply ongoing context to AI agents. Context engineering is the discipline; a Context OS is one kind of system that operationalizes it.
Last updated:
Origins
Where the term comes from
The term gained currency through LangChain's blog and engineering content in 2024-2025, then was reinforced by Anthropic's engineering documentation on building effective agents. The core observation: getting the right context to an LLM at the right time is the highest-leverage skill in production AI work, often more important than prompt wording or model choice.
Capabilities
What Context Engineering does
Retrieval design
Decides what context to pull from where, when, and at what granularity. Includes RAG, graph traversal, and tool-based retrieval patterns.
Context window management
Manages what fits in the LLM's context window: which pieces of context to include, what to compact, and what to drop.
Prompt structure
Decides how retrieved context is laid out in the prompt: ordering, formatting, signposting, and explicit attribution.
Ongoing context supply
Systems and protocols (like MCP) that supply context to AI agents continuously, not just per-prompt.
Distinctions
Context Engineering vs adjacent concepts
Context Engineering is often confused with related but distinct ideas. Here is how it differs.
| Concept | What it is | How Context Engineering differs |
|---|---|---|
| Prompt engineering | Crafting the wording of a one-shot prompt. | Systematizing how context flows into every prompt across an agent's lifecycle. |
| RAG | One mechanism for fetching context. | A broader discipline that includes RAG, plus retrieval design, context budget management, and tool integration. |
| Context OS | The discipline. | The system that operationalizes the discipline at company scale: the source layer that supplies context to every AI agent. |
Who uses it
Who uses Context Engineering
AI engineers and ML platform teams building production AI systems. Founder-operators evaluating how their company knowledge gets to AI tools. Anyone designing AI agents that need ongoing, grounded context across many tasks.
FAQ
Common questions about Context Engineering
Is context engineering the same as prompt engineering?
No. Prompt engineering is about wording an individual prompt. Context engineering is about systematically delivering the right context to AI agents over time, across many prompts and many tasks. Prompt engineering is one tactic within context engineering.
What systems does context engineering involve?
Retrieval systems (RAG, graph traversal, full-text search), context window managers (compaction, ordering, budget allocation), tool servers (MCP, function calling), source layers (knowledge graphs, document stores), and observability tools that track what context an agent saw and why.
How does a Context OS relate to context engineering?
A Context OS is the source layer in a context engineering setup. It is the structured, queryable knowledge that retrieval systems pull from. Context engineering is the discipline that decides what gets pulled, when, and how it shows up in the prompt.
Related terms
Keep going
See Context Engineering in practice
DearTech-OS is a Context OS for founder-operators. Explore the product or talk through whether one is right for your team.