Certifications
Anthropic runs a certification program for Claude with four exams — an associate, a developer, and two architect levels. This page is the map: what each exam is for, which one you should actually sit, and which pages of this course cover each exam's material. Then take the Exam Simulator to find out whether you're ready.
There are four badges you can earn for knowing Claude. One is for people who explain it to others, one is for people who write the code, and two are for people who design the whole system — a starter one and a harder one. They cost money and you sit them at a test centre or at home with someone watching over the webcam, so it's worth knowing which one you actually want before you book.
This course is not affiliated with, endorsed by, or an official product of Anthropic. "Claude", the certification names, and exam codes are Anthropic's. Exam fees, formats, and policies change — always confirm current details on Anthropic's own certification pages and at Pearson VUE before you book. Nothing here reproduces real exam questions; that would breach every exam's terms and would teach you nothing.
The four exams
☺ Like you're 10: Four badges. Pick the row that sounds like your job.
| Code | Name | Who it's for | Fee |
|---|---|---|---|
| CCAO-F | Claude Certified Associate — Foundations | Non-engineers who need to speak about Claude credibly: solution consultants, PMs, enablement, pre-sales. Capabilities, use cases and adoption patterns rather than code. | $99 |
| CCDV-F | Claude Certified Developer — Foundations | The hands-on engineering exam. The Messages API, the Agent SDK, MCP servers, prompt and context engineering, evals, cost control and security. | $125 |
| CCAR-F | Claude Certified Architect — Foundations | People who design the systems other people build — agentic architecture, tool design, reliability, and the trade-offs between them. | $125 |
| CCAR-P | Claude Certified Architect — Professional | The step up from CCAR-F: harder scenarios, deeper judgement calls, production-scale constraints. | $175 |
Format and booking
☺ Like you're 10: Two hours, someone watches you take it, and you get a score out of 1000 with 720 to pass.
| Detail | What to expect |
|---|---|
| Length | 120 minutes per exam. |
| Proctoring | Proctored. Delivery moved to Pearson VUE in mid-2026, including Pearson OnVUE for online-proctored sittings from home. |
| Scoring | Scaled score on a 100–1000 range; 720 to pass. A scaled score is not a percentage — don't reason about it as "72% of questions right". |
| Registration | Through the Anthropic Partner Academy, which also hosts self-paced preparation material. |
| Retakes | 14 days after a first failure, 30 days after a second, 90 days after a third, with a maximum of four attempts on the same exam in any 12-month period. Plan accordingly — a rushed second attempt costs you the whole window. |
Which one should you take?
☺ Like you're 10: Take the one that matches the job you actually do — not the hardest one.
- You explain Claude to other people. CCAO-F. It rewards knowing where Claude fits and where it doesn't, which is a different skill from building with it.
- You write the integration. CCDV-F. This course's Foundations and Agents tracks are aimed squarely at it.
- You choose the shape of the system. CCAR-F, then CCAR-P once you've run something real in production. Sitting CCAR-P without production scars is the common mistake.
- You want one credential and you write code. CCDV-F. It's the most directly useful, and it's the one whose material you'll use the day after.
A certification is evidence of a floor, not a ceiling. It proves you know the vocabulary and the mechanics well enough to be trusted with a task — it does not prove you've shipped anything. The capstone projects in this course are the part that does.
How this course maps onto the exams
☺ Like you're 10: Which pages to read for which badge.
| Exam area | Read these | Mostly for |
|---|---|---|
| What Claude is, capabilities and limits | What Is Claude?, Vision & Multimodal | CCAO-F |
| Messages API mechanics | Messages API, Streaming & Context, Structured Outputs | CCDV-F |
| Prompt & context engineering | Prompting Fundamentals, Advanced Prompting, Extended Thinking | All four |
| Tool use and MCP | Tool Use, Agent SDK & MCP | CCDV-F, CCAR-F |
| Agentic architecture | Building Effective Agents, Patterns & Anti-Patterns | CCAR-F, CCAR-P |
| Retrieval | RAG with Claude | CCDV-F, CCAR-F |
| Cost, caching and latency | Prompt Caching, Production Deployment | CCDV-F, CCAR-P |
| Evaluation | Evaluation & Testing | CCDV-F, CCAR-P |
| Safety and security | Safety & Responsible Use | All four |
| Developer tooling | Setting Up Your Environment, Claude Code | CCDV-F |
A study order that works
☺ Like you're 10: Read, then test yourself, then find out what you got wrong — in that order.
- Work the Foundations in order. Don't skip the prompting pages because they look easy; they carry the most exam weight of anything here.
- Build one capstone. The scenario questions on the architect exams are much easier to read once you've made these decisions for real.
- Run the flashcards and the glossary until the vocabulary is automatic — scenario questions are slow to read when the terms aren't.
- Sit the Exam Simulator under real conditions: timer on, no notes, one sitting.
- Go back to the pages behind every question you missed. Re-sit. Book the real thing when the simulator stops surprising you.
Two questions before you book anything. First: which exam matches the work you do this quarter — not the one with the most impressive name? Second: what's the pass mark, and why is it not a percentage? If either answer is fuzzy, re-read the two tables above before you spend the fee.
Check your answers
- Which exam: CCAO-F if you explain and position Claude; CCDV-F if you build with it; CCAR-F then CCAR-P if you design systems others implement. Most engineers reading this course want CCDV-F.
- The pass mark: 720 on a scaled 100–1000 range. Scaled scoring normalises across different question sets, so 720 does not mean "72% correct" — it means a fixed standard of competence, however hard your particular set happened to be.