/ Directory / Playground / Anthropic Courses
← All servers ↗ GitHub
● Official anthropics ⚡ Instant

Anthropic Courses

by anthropics · anthropics/courses

Anthropic's official Claude curriculum — prompt engineering, tool use, RAG, and evals — packaged as a learnable skill resource.

anthropics/courses is the official Anthropic curriculum for working with Claude effectively. Includes the API fundamentals course, prompt engineering, real-world prompting, prompt evaluations, tool use, and a customer-service agent walkthrough. Use it as a SKILL resource to teach Claude how Claude is best used — meta but high-leverage.

▶ Watch demo 📦 Install 💡 Use Cases

Why use it

Key features

Live Demo

What it looks like in practice

ready

Install

Author repository · Claude Skills documentation

Use Cases

Real-world ways to use Anthropic Courses

Level up your prompt engineering using the official curriculum

👤 Devs whose prompts feel hit-or-miss ⏱ ~60 min intermediate

When to use: You want fundamentals not folklore.

Flow
  1. Index the course
    Use anthropic-courses. Summarize the prompt engineering chapter outlines.✓ Copied
    → Section index
  2. Apply to my problem
    Here's a prompt I keep tuning unsuccessfully. Apply the chapter's techniques to rewrite it.✓ Copied
    → Rewritten prompt with citations to lessons
  3. Eval the result
    Build a small eval harness from the evaluations chapter to compare old vs new prompt.✓ Copied
    → Eval table showing improvement

Outcome: Better prompts, grounded in Anthropic's own teaching.

Implement tool use following the official walkthrough

👤 Devs first integrating tool calling ⏱ ~45 min intermediate

When to use: Docs feel abstract; you want a working example.

Flow
  1. Lift the template
    Use anthropic-courses tool-use chapter. Lift the basic example into my repo and adapt.✓ Copied
    → Working tool-use loop in your code
  2. Add my actual tools
    Replace the example tools with ones for my domain (search-orders, refund-customer).✓ Copied
    → Domain-specific tool defs

Outcome: Tool use shipped fast with official patterns.

Build a real eval harness following the prompt-evaluations course

👤 Teams shipping LLM features ⏱ ~90 min advanced

When to use: You're shipping prompts and want regression coverage.

Flow
  1. Lift the harness
    Use anthropic-courses prompt-evaluations chapter. Set up the harness for my prompt.✓ Copied
    → Eval harness running
  2. Write 20 cases
    Generate 20 representative + edge cases. Score expected outputs.✓ Copied
    → Test set ready
  3. Wire to CI
    Run on every prompt change in CI. Fail if accuracy drops >2%.✓ Copied
    → CI gate

Outcome: LLM features that don't silently regress.

Combinations

Pair with other MCPs for X10 leverage

anthropic-courses-skill + anthropic-skills-official

Learn fundamentals + ship in skill format

Apply the prompt-engineering lessons inside an anthropics/skills-formatted skill.✓ Copied
anthropic-courses-skill + anthropic-cookbook-skill

Curriculum + cookbook = theory + practice

Read course chapter, then look up matching cookbook recipe to ship faster.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
(notebook curriculum) natural lookup When learning a new Claude technique 0

Cost & Limits

What this costs to run

API quota
N/A — local
Tokens per call
Course content can be long — pin to a specific chapter
Monetary
Free (MIT)
Tip
Use as a reference, not a context dump — let Claude search relevant sections

Security

Permissions, secrets, blast radius

Credential storage: None
Data egress: None

Troubleshooting

Common errors and fixes

Notebooks won't render

Skill is for reference, not execution; open notebooks in Jupyter or VS Code

Course examples use older API patterns

Skim the most recent commits; Anthropic updates the curriculum but legacy patterns linger

Alternatives

Anthropic Courses vs others

AlternativeWhen to use it insteadTradeoff
anthropics/anthropic-cookbookYou want recipes, not lessonsCookbook is action-oriented; courses are foundational
anthropics/anthropic-quickstartsYou want runnable templates to forkQuickstart = boilerplate; courses = pedagogy

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills

Continue exploring

Related tools and workflows

Choose another guide for your workflow, or explore a collection.

Browse all topic collections → · See 29 new official additions →