Claude · Module 07 of 8

Advanced

Context engineering, tool use, and multi-agent patterns for power users.

2 lessons8 exercises
Topics & techniques

Context engineering

Control what Claude sees, not just what you ask.

4 cards →
Concept
Curate the Context — Don't Just Dump It
Context engineering: deliberately choosing what Claude sees — the relevant files, facts, and constraints — rather than pasting everything and hoping the model finds the signal.
L4 · Multi-AI systems
Workflow
Give Claude a Memory File for Long Projects
A pattern for work that spans many sessions: keep a single, hand-curated 'memory' document — key decisions, constraints, current state, open questions — and paste it (or attach it) at the start of each new chat. It's the durable spine the context window can't provide on its own.
L4 · Multi-AI systems
Best practice
Put the Most Important Context Where Claude Looks
A placement habit for long prompts: put the material that most determines the answer at the start or the very end of the context, and let bulk reference material sit in the middle. Order is not cosmetic — where a fact sits changes how strongly it lands.
L4 · Multi-AI systems
Best practice
Strip the Context Down to What Changes the Answer
A discipline of removing context that doesn't change the answer before you send the prompt. The test for every file, paragraph, or example is simple: if I delete this, would a correct answer change? If not, it's noise — cut it.
L4 · Multi-AI systems

Tool use

Give Claude tools and functions to call — and design them so it calls them well.

4 cards →
Best practice
Describe Tools Like You'd Brief a New Hire
A tool-design principle for function calling: write each tool's description and parameter docs as if briefing a capable new hire who's never seen your system — what it does, when to use it (and when not), what each argument means, and what happens on failure — rather than a terse one-liner.
L4 · Multi-AI systems
Best practice
Give Claude Fewer, Sharper Tools
A tool-set design principle: expose the smallest set of tools that covers the job, each doing one clear thing, rather than a sprawling menu with overlapping or near-duplicate functions. Tool accuracy is a selection problem before it's a calling problem.
L4 · Multi-AI systems
Best practice
Make Tools Fail in a Way Claude Can Recover From
A design rule for tools Claude calls: when a tool fails, return an error the model can act on — what went wrong, why, and what a valid call would look like — instead of a raw stack trace or a bare status code. The error message is part of the prompt Claude reads next.
L4 · Multi-AI systems
Best practice
Get Structured Output You Can Trust, Not Just JSON-Shaped Text
A practice for getting machine-readable output: define the exact schema you need, ask Claude to produce only that, and then validate the result against the schema in your code — treating the model's output as untrusted input to be checked, not as guaranteed-valid data.
L4 · Multi-AI systems
Advanced Claude techniques

Context engineering is the real advanced skill

A large context window tempts people to dump everything in, but the advanced move is the opposite: curate what Claude sees so the signal is not buried in noise. Context engineering means giving Claude the slice that matters — the relevant files, the key constraints, the pertinent history — and leaving out the rest. Done well, it is the single biggest lever on output quality, more than any phrasing trick. The window is a tool to be filled deliberately, not a bin to be filled fully.

Tool use turns Claude into an agent

The other advanced frontier is tool use: giving Claude functions it can call to fetch data, run actions, or interact with your systems. The craft is in the interface — describe each tool the way you would brief a new hire, with a clear name, purpose, and inputs, and ask Claude to return structured output you can actually parse. Get that right and Claude stops being a text generator and becomes a component you can wire into real software.

The advanced topics to explore

This section covers context engineering (curating what the model sees) and tool use (giving Claude functions and structured output). Work through the cards below once the fundamentals are automatic and you want production-grade control.

Frequently asked
What is context engineering in Claude?

Context engineering is deliberately curating what you put in Claude's context window — the right files, constraints, and history — so the relevant signal is not diluted by noise. It is the highest-leverage advanced skill, ahead of any prompt-wording trick.

How does tool use work with Claude?

You give Claude a set of functions it can call, each described with a clear name, purpose, and inputs, and ask it to return structured output. Done well, this lets Claude fetch data and take actions, turning it into a component you can build software around.

Does filling Claude's whole context window help?

No — including irrelevant material buries the parts that matter and can degrade the answer. The goal is the right context, not the most context, which is why context engineering matters even with a large window.

How do I get structured output from Claude?

Specify the exact shape you want and describe your tools clearly, so Claude returns data you can parse reliably rather than free-form prose. The tool-use cards below cover the patterns that make output dependable.

Test yourself

Have you mastered advanced Claude technique?

Six questions on context engineering and tool use — what to include, where to put it, and how to design tools Claude calls well. No sign-up, instant score, and you can share how you did.

  1. Q1 What's the test for whether a file belongs in your context?
  2. Q2 You have a long prompt: a key constraint, the question, and bulk reference material. Where does the constraint go?
  3. Q3 You're working across many sessions on one project. What keeps Claude up to speed?
  4. Q4 How should you write a tool's description for Claude?
  5. Q5 Your tool fails. What should it return?
  6. Q6 Claude returns JSON matching your schema. What now?
0 / 6 answered