All learning paths

AI learning path

Context engineering and MCP

Design the information and tool boundary around an agent instead of stuffing everything into a prompt.

Best for
Developers using coding agents and tool-connected workflows
Level
Intermediate to advanced
Time
10-18 hours

Choose this when

You are using coding agents, custom tools, MCP servers, skills, memories, or connected workspaces.

You should be able to

You can decide what belongs in instructions, retrieval, tools, MCP, subagents, or project files.

Checkpoint

Move on when you can remove context without making the agent worse.

Do

Learning sequence

This is the route through the topic. Watch and open the material inside the step where it is used.

Search resources Search videos

Step 1

Separate context types

Classify what should be durable instruction, retrieved knowledge, live tool data, or temporary conversation.

  • Instructions
  • Retrieval
  • Tools

Watch here

Claude Skills, Clearly Explained

Peter Yang

Introduces reusable skills and context packaging for agent workflows.

Open here

Model Context Protocol Tutorial

Free tutorial · Matt Pocock · Intermediate

You want to understand MCP and build TypeScript MCP servers over stdio or HTTP, connect Claude Code to tools, use MCP prompts, and package servers for distribution.

Open resource

Step 2

Expose tools deliberately

Use MCP or APIs for actions and live data where a text dump would be brittle.

  • MCP servers
  • Permissions
  • Schemas

Watch here

Open Questions for AI Engineering

Simon Willison

Useful for thinking about tool boundaries and agent context design.

Open here

Step 3

Keep it testable

Create small repeatable tasks that reveal whether the context design actually improves agent behavior.

  • Fixtures
  • Eval tasks
  • Context budgets

Open here

Hugging Face MCP Course

Free course · Hugging Face · Beginner to intermediate

You want a free structured MCP path with concepts, assignments, SDKs, and a certificate route.

Open resource

Model Context Protocol intro

Spec and guide · Model Context Protocol · Intermediate

You need the core mental model for how tools, connectors, and agent environments interoperate across AI apps.

Open resource

Practice task

Refactor one overloaded agent prompt into project instructions, a small resource file, and one tool or MCP boundary.

Reference

All resources in this path

Search resources

Step 1

Model Context Protocol Tutorial

Free tutorial · Matt Pocock · Intermediate

You want to understand MCP and build TypeScript MCP servers over stdio or HTTP, connect Claude Code to tools, use MCP prompts, and package servers for distribution.

Step 2

Hugging Face Context Course

Free course · Hugging Face · Intermediate

You want a focused course on context engineering for code agents, including skills, subagents, hooks, and MCP.

Step 3

Hugging Face MCP Course

Free course · Hugging Face · Beginner to intermediate

You want a free structured MCP path with concepts, assignments, SDKs, and a certificate route.

Step 3

Model Context Protocol intro

Spec and guide · Model Context Protocol · Intermediate

You need the core mental model for how tools, connectors, and agent environments interoperate across AI apps.

Educators to follow

Matt Pocock profile photo

Matt Pocock

Beginner to advanced

Use LLM Fundamentals or the AI Engineer Roadmap if you need concepts, the Vercel AI SDK Tutorial or AI SDK v6 Crash Course if you want to build apps, and the AI Skills catalog if you want practical agent workflows like /teach, /grill-me, /tdd, and /triage.

View educator
Simon Willison profile photo

Simon Willison

Beginner to advanced

Read the recent model-roundup posts, then try the llm command-line tool with two or three different providers.

View educator
Josh Pigford profile photo

Josh Pigford

Beginner to intermediate

Read the public notes and examples before deciding whether the paid material matches your business.

View educator