The Context
Engineering Guide
Prompt engineering tells AI what to do. Context engineering tells AI who it is.
New to Claude Code? Start with the free mini tutorial first.
What's inside
Audit your AI setup in 60 seconds
The guide explains the theory. The Health Check MCP scores your actual setup across 3 dimensions — so you know exactly what to fix first.
Get the free guide
Instant access. No fluff.
- Complete guide
- The 7 CE patterns explained
- Health Check MCP audit tool
No spam ever.
Keep context current, specific, and useful.
AI context management is the work of deciding what an AI system should know for a task, where that information lives, and how it is corrected when it changes. The five layers below separate stable instructions from working memory, connected tools, repeatable skills, and automatic checks.
| Layer | Use it for | Keep it useful by |
|---|---|---|
| CLAUDE.md | Stable instructions and routing | Pointing to the smallest relevant source instead of repeating everything. |
| Memory | Decisions, preferences, and lessons that survive a session | Recording source, owner, and the moment a fact was last confirmed. |
| MCP | Retrieval and actions across the tools where work happens | Keeping permissions and tool scope explicit. |
| Skills | Repeatable work with a defined input, output, and review | Updating the instruction when a correction repeats. |
| Hooks | Checks that run at the right point in a workflow | Making the trigger and expected evidence visible. |
When a question depends on relationships between sources, owners, commitments, and changes, govern AI context with evidence, freshness, and corrections. If you want this operating layer in your own work, build this context system in your Second Brain, then add a durable memory layer for the knowledge that needs to last.
From the Blog
Go deeper with these related articles.
I Run a $48K Business With Zero Employees
The full case study: 30 systems, 12 MCP connections, one terminal. Every system I built.
ReadContext Engineering with MCP
Five production-tested patterns from running 12 MCP servers daily across 100+ client deployments.
ReadThe File System Is the Prompt
The 5-layer context engineering model: CLAUDE.md, Memory, MCP, Skills, Hooks.
Read20 Papers on Context Engineering
The evidence base behind context engineering - what research actually says works.
ReadFrom Prompt Engineer to Context Architect
Why context architecture is becoming the most important career skill of 2026.
ReadWho This Guide Is For
- Knowledge workers using Claude Code daily who want to stop re-explaining context every session
- People who've tried AI tools but aren't seeing compounding results — just the same effort every time
- Anyone frustrated that their AI "forgets" between sessions and can't build on past work
- People ready to move beyond prompts and build a system that gets smarter the longer you use it
Iwo Szapar
"(A)I will help you work smarter, not harder."
- •Co-founded the AI Maturity Index (acquired by ISG in 2026)
- •Led training & change management projects at Microsoft, ING Bank, Walmart
- •Advised Saudi Ministry of Communications on AI up-skilling
- •Up-skilled 25,000+ professionals via Remote-how Academy
Trusted By Teams At
Get the Free Guide
From zero to a compounding AI system.
Ready to go deeper?
The guide explains the framework. Two products build it for you.
The complete Claude Code infrastructure — CLAUDE.md, 30+ agents, 20+ skills, progressive disclosure, and hooks. Every pattern from this guide, pre-configured.
Starter $197 · Accelerator $597 · Launch Partner $2,497
See Second Brain 2.0The memory layer that makes the whole system compound. Personalized to your role, tools, and workflow. Patterns extracted automatically.
Free · Standard $199/yr · Pro $349/yr
Get MemoryOS