CertifyChange track
Claude Certified Architect – Professional
0%
Plan not started
Blueprint · a Claude academy
Learn to build with Claude.
Prove it with a certification.
Hands-on projects that teach the patterns teams actually ship, and structured tracks for every Claude certification. Free, self-paced, and it reads itself aloud.
Loading progress…
Build
Real Claude solutions, end to end
16 hands-on projects: support copilots, document pipelines, text-to-SQL, agents with approvals, MCP servers, eval harnesses, Claude Code rollouts and more, plus Beyond Claude builds on OpenAI, LangChain, LangGraph, LiteLLM and local models so you see the whole landscape. Each has a step-by-step guide and a tested solution you can download.
This week
Chapters done
–
Chapter quizzes
–
Projects built
–
Days to exam
–
Course
Week-by-week plan
How to study with this plan
You're a senior architect, so a lot of this will feel familiar. What the exam tests is whether you pick the Anthropic-recommended answer under constraints. Budget about 10–11 hours a week and work through each chapter in the same order every time.
For each chapter (≈2–3 h)
- Lesson tab: plain idea, analogy, example. 15 min
- Practice Qs cold, before studying. Don't worry about the score; it shows you what to look for. 10 min
- Deep dive: read the walkthrough, trace the worked case, run or read the code. 30–40 min
- Study notes: tables and exam traps; add your own takeaways in My notes. 25 min
- Read the linked Anthropic docs/posts. The exam paraphrases them. 20–30 min
- Lab: build or sketch it. This is where it sticks. 30–45 min
- Practice Qs again, then rate confidence 1–5 and mark Done. 10 min
A typical week
- Mon–Thu, 60–75 min/day: one chapter's Lesson, cold Qs, Deep dive and Notes.
- Sat, ~3 h: labs for the week's chapters, plus the linked readings.
- Sun, ~2 h: practice mode on this week's domain (15–25 Qs), then the “Review due” list.
- Every day, 10 min: the review-due items in the summary panel (spaced repetition).
Behind? Skip labs for chapters you've already built in real life, never the practice questions.
Is this enough?
- This portal: your structured course, notes and 198 practice questions.
- Official exam guide (must): read it at the start and again in week 6; tick every objective.
- Anthropic Academy prep path: if you have Partner Academy access, do it alongside the course weeks.
- Anthropic docs and engineering posts: linked in each chapter. They're the source most questions come from.
- Hands-on: at least the labs for 1.3, 2.4, 3.1, 3.4, 4.2 and 7.1. Real usage beats reading.
- Blueprint full mocks: Practice → Full mock exam, timed and weighted like the real thing. Take two in the final week.
About 10–11 hours a week. Open a chapter and work through its tabs; your checklist fills in as you go. When you're finished, press ✓ Mark chapter complete at the bottom of the chapter. Turn on “Hide completed chapters” to see only what's left.
Hands-on projects
Build it to own it
Scenario questions are much easier once you've built the thing. These 16 projects cover the typical ways organisations use Claude, from support bots and document pipelines to agents, MCP servers, evals and Claude Code rollouts. The Beyond Claude projects (P17–P23) rebuild the same ideas on OpenAI, the OpenAI Agents SDK, LangChain, LangGraph, LiteLLM and local Ollama models, so you can compare ecosystems and talk about them with confidence. Each one has an overview, a detailed step-by-step guide with the exact commands and what you should see, and a complete working solution you can read, copy or download. New to building? Start with Project 0 just below. Aim for at least one project per week; the recommended week matches the course.
Setup once, reuse everywhere
LanguagePython 3.11+, the
anthropic SDK, python-dotenv for your API keyDataDuckDB locally (your Snowflake instincts transfer), Parquet files, SQLite for mock systems
RetrievalLanceDB or Chroma for vectors,
rank_bm25 for keywords, an embeddings provider such as Voyage AI or a local sentence-transformers modelAgents and MCPClaude Code, the Claude Agent SDK, and the MCP Python SDK (FastMCP)
Quality
pytest, sqlglot for SQL validation, a JSONL eval format you reuse from P14Beyond ClaudeP17–P23 add the
openai SDK and Agents SDK, LangChain, LangGraph, LiteLLM and Ollama: one extra pip install -r requirements-beyond.txtCost controlSet a monthly spend limit in the Console; use Haiku and the Batch API while iterating
Full mock exam
Same length, time limit and domain weighting as the real exam, drawn from Blueprint's question bank. The score is an estimate on the exam's 100–1000 scale (720 ≈ 72% correct) and goes straight into your mock exam log.
Practice mode
Draw a random set from the 198-question bank, filtered by domain. Questions follow the exam's style: scenario-based, single answer or "Select TWO". Your results are saved so you can see which domains need work.
| Date | Filter | Score | By domain |
|---|
Chapters in progress
0
Avg confidence
–
Best mock
–
Practice accuracy
–
Domains by exam weight
The faint band is each domain's share of the exam; the solid bar is how much of that domain you've finished. The exam weight tells you where points come from.
Syllabus updates
Changes your admin has applied since this track's guide was first loaded. Updated chapters also show a note at the top of their lesson.
Mock exam log
Log your practice scores on the 100–1000 scale (or convert a percentage: roughly 72% ≈ 720). Aim for two mocks at 800+ before you book.
| Date | Source | Score | Result | Weakest |
|---|
Glossary
Every key term from the deep dives, searchable. Good for a 10-minute review before bed.
Core sources
- Anthropic Partner Academy prep path and certification page (registration runs through here)
- Anthropic Engineering blog — most scenario questions echo these posts
- Claude Docs, Claude Code docs, MCP spec
Certification tracks
Pick your track
Blueprint covers every Claude certification. Each track has its own weekly plan, chapters, practice bank and projects, all weighted to that exam's official domains. Your progress is saved separately for each track, so switching never loses anything.
Signed in as
Guest
Your account
No account needed to study. Sign in to keep your progress safe and pick it up on any device; whatever you've done as a guest comes with you.
Your team
You're not in a team yet. Your admin can give you a team code so your progress shows up in their team view.
What gets saved where
- Your progress, notes and scores are private to you: stored under your account, one copy per track.
- Admins see a summary only: chapters done, quiz and practice accuracy, projects built and last active date. Never your notes or answers.
- Not signed in, or opened with view-only access? Progress stays in this browser.
Start this track over
Clears chapters, quizzes, practice history, mocks and projects for this track only. Other tracks aren't touched.
Configuration · admins only
Run the academy
Who's studying, how they're doing, the teams and tracks on offer, and the syllabus itself. Everything here is shared: changes apply to every learner the moment you save.
By track
| Track | Learners | Active 7d | Avg done | Avg quiz | Status |
|---|
| Learner | Track | Team | Done | Quiz | Practice | Projects | Best mock | Exam | Last active |
|---|
People appear here after they open Blueprint while signed in. Summaries only: nobody's notes or answers are visible to admins.
| Team | Join code | Members | Avg done |
|---|
Share the join code with the team. Learners enter it under Tracks → Your team.
Tracks offered
Hidden tracks disappear from the picker for learners. People already on a hidden track keep their progress.
Announcement banner
Shown at the top of every page until the learner dismisses it. Keep it to a sentence or two.
Check the official syllabus
Starts a Claude research run that reads Anthropic's current exam guides and compares them with what Blueprint teaches: exam facts, domain weights, objectives, and anything a chapter now needs to mention. It never changes the site by itself. Each difference comes back below as a proposed update for you to apply or reject.
Settings
The check runs as a scheduled task in the account that owns Blueprint. Only admins whose Claude account can run that task can start it from here; everyone else can copy the request and paste it into a Claude chat.
Import updates by hand
Paste a JSON array of updates (same shape the checker writes). They arrive as pending so you can review them first.
Current overrides
What's been applied on top of each track's built-in guide. Resetting a track puts it back to the built-in version.
Backup
Downloads a JSON file with the shared configuration: tracks, teams, syllabus overrides, update history and the roster summary. Private learner progress isn't included; only each learner can read it.
Tidy up
Removes applied and rejected updates older than 30 days. Pending ones are kept.
About this install
Margin Notes
Things you flagged to revisit
Select any text while you read a chapter and tag it Revisit, Confusing or Exam trap. It stays highlighted in the chapter and shows up here with your note. Tick each one off as you review it; aim to clear the list in your final week.