60-Day Comprehensive Plan (CCAR-P)¶
Level: Intermediate to Advanced · Time: 1.5 to 2 hours a day, ~110 hours total · Track: CCAR-P (Claude Certified Architect – Professional)
Who this plan is for¶
This plan is for experienced technical professionals: software engineers, platform engineers, solution or enterprise architects, security engineers and technical leads with several years of experience who want to prepare thoroughly for CCAR-P. It assumes you can read Python, understand APIs, authentication and cloud architecture, and ideally have worked on at least one project that used an LLM.
It is not suitable for complete beginners, either to technology or to AI at speed. If you are a developer who has never worked with LLMs, take the 90-day enterprise engineering plan, which builds the foundations first. If you are not technical, the CCAO-F plans (30-day, 15-day) are the right starting point.
The exam guide recommends 3+ years in systems architecture or platform engineering and 6+ months hands-on with Claude or comparable LLM systems in production; the format and domain weights are recorded in registry.yaml and summarised in Exam facts. Hands-on labs are the substitute for some of that production experience, so do not skip them.
gantt
title 60-day CCAR-P plan
dateFormat X
axisFormat Day %s
section Build knowledge
Foundations and API :1, 8
Prompting and context (D2) :8, 15
Tools and MCP (D3) :15, 22
RAG and agents (D1, D3) :22, 29
Security and governance (D5) :29, 36
Evaluation and operations (D4) :36, 43
Architecture, stakeholders, Claude Code (D1, D6, D7) :43, 50
section Apply and test
Project and gap repair :50, 57
Final revision :57, 61How to use this plan¶
- Numbers in brackets in the Objectives column are CCAR-P objective IDs: "2.2" means
CCAR-P-2.2in the registry. - Each row is one day. Weekends can absorb a missed day; the seventh day of each week is a revision milestone.
- Labs run free in mock mode (environment setup). Live mode is optional.
Cost: live-mode lab runs require API credits. Set a spend limit in the Console first.
- Practice questions and mocks run in the mock exam engine; use
--track CCAR-P --domain Nfor domain practice. - Keep a study log of every missed checkpoint and the page that answers it.
- Skim pages you already know well, but always answer their "Check your understanding" questions.
Week 1 (days 1 to 7): foundations refresh and the Claude API¶
Focus: Domains 2 and 3 groundwork.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 1 | Refresh LLM fundamentals: tokens, context windows, hallucination (CCAR-P-2.4, 5.2) | Large language models · Tokens · Context windows · Hallucinations | Lab 01: First Claude interaction in mock mode; read the lab code and the shared helper | 1.5 h | All four page quizzes correct; explain why history re-sending drives cost |
| 2 | Explain Claude's training, statelessness and roles; Anthropic's safety layers (2.2, 5.1) | How Claude works · Anthropic's safety approach · Responsible use | Draft a one-paragraph explanation of the Usage Policy's high-risk requirements for an engineering audience | 1.5 h | Name the constitution's four priorities and the RSP's core idea |
| 3 | Map the product ecosystem; select models on trade-offs (2.1, 7.1) | Product ecosystem · Model families · Model selection | Lab 06: Model comparison; compare two models at two effort levels | 2 h | Justify a model and effort choice for three workloads |
| 4 | Make Messages API calls; handle stop reasons (3.3) | Getting started with the API · Messages API | Lab 07: API integration, first half | 2 h | Write from memory a request with system prompt, messages and a stop_reason check |
| 5 | Use streaming, handle errors and retries, count tokens (3.3, 4.5) | Streaming · Errors, rate limits, retries · Token counting and cost | Lab 07: API integration, second half | 2 h | Classify five API errors as retryable or not |
| 6 | Use documents, citations and batches; know cloud platform differences (3.7) | Vision, documents and citations · Batch processing · Cloud platforms | Sketch when you would choose the Batch API and when a cloud platform; note feature-availability caveats | 1.5 h | Explain one advantage of each cloud platform option |
| 7 | Revision milestone 1 | Pages linked from any missed quiz this week | Fundamentals flashcards; a 15-question learning-mode drill on CCAO-F Domain 3 in the mock exam engine (python 18-mock-examinations/run_mock_exam.py --track CCAO-F --domain 3 --mode learn --count 15, or the same in the browser app) |
1 h | 75% or more on the Domain 3 drill |
Week 2 (days 8 to 14): prompting and context engineering¶
Focus: Domain 2.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 8 | Write clear, structured prompts with examples (2.2, 2.3) | Clear, direct instructions · XML tags · Few-shot examples | Lab 02: Prompt improvement | 1.5 h | Rewrite a weak system prompt into a structured one |
| 9 | Apply reasoning, decomposition and structured outputs (2.3) | Reasoning and thinking · Task decomposition and chaining · Structured output | Lab 03: Structured outputs | 2 h | Explain when to chain prompts versus one prompt |
| 10 | Design system prompts, templates and guardrails; iterate systematically (2.2) | System prompts and roles · Iterating on prompts · CCAR-P study guide D2 | Write a production system prompt for a fictional support assistant with scope, tone and refusal guidance | 2 h | Peer-review checklist for system prompts from memory |
| 11 | Manage long context and grounding (2.4) | What is context engineering? · Managing long context | Lab 04: Document-grounded Q&A | 1.5 h | List three tactics against lost-in-the-middle |
| 12 | Use prompt caching and compaction to optimise tokens (2.4, 2.5, 4.5) | Prompt caching · Compaction and context editing | Lab 05: Context management · Lab 17: Cost optimisation | 2 h | Lay out a request for maximum cache hits and say how to verify them |
| 13 | Design memory, persistence, progressive disclosure and Skills (2.5, 3.8) | Memory and persistence · Progressive disclosure and Skills · Projects and knowledge | Compare a monolithic context design with a progressive-discovery design for a 300-document knowledge base | 1.5 h | Explain progressive discovery versus monolithic context in two sentences |
| 14 | Revision milestone 2 | CCAR-P study guide D2 (re-read weak sections) | Domain 2 practice questions in the mock exam engine, learning mode | 1.5 h | 80% or more on Domain 2 |
Week 3 (days 15 to 21): tools, MCP and integration mechanisms¶
Focus: Domain 3.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 15 | Define tools and run the tool-use loop (3.1) | Tool use fundamentals · Defining tools · The tool-use loop | Lab 08: Tool invocation | 2 h | Trace a two-tool conversation, message by message |
| 16 | Use server tools; design and secure tools; avoid capability bloat (3.1, 3.2) | Server tools · Tool design best practices · Tool security | Audit a fictional agent with 25 tools and propose a smaller tool surface | 2 h | Name three signs of capability bloat |
| 17 | Explain MCP and its architecture (3.7) | What is MCP? · MCP architecture | Diagram host, client and server for a fictional integration | 1.5 h | Explain MCP's roles and transports |
| 18 | Build and use an MCP server (3.7) | Building an MCP server · Using MCP with Claude | Lab 09: MCP fundamentals | 2 h | Explain how a remote MCP server is authorised |
| 19 | Secure MCP; compare MCP with API, CLI and agent-to-agent options (3.2, 3.7) | MCP security · MCP versus alternatives · MCP server security | Write a short decision record choosing an integration mechanism for one scenario | 2 h | Pick an integration mechanism for four scenarios |
| 20 | Analyse authentication and authorisation in app-level integrations (3.2) | Connectors and integrations · Agent identity and permissions · CCAR-P study guide D3 | Map permissions for a connector rollout: who, what, read or write, approvals | 1.5 h | Explain the risk of an over-privileged service identity |
| 21 | Revision milestone 3 | Pages linked from missed questions | Domain 3 practice questions (tools and MCP objectives) | 1.5 h | 75% or more |
Week 4 (days 22 to 28): retrieval and agentic architecture¶
Focus: Domains 1 and 3.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 22 | Explain RAG; choose chunking strategies (3.5) | RAG fundamentals · Chunking strategies | Chunk one long fictional policy three ways and compare | 1.5 h | Choose a chunking strategy for three document types |
| 23 | Choose embeddings, indexing and retrieval strategies (3.5, 3.6) | Embeddings and indexing · Retrieval strategies | Lab 11: RAG pipeline | 2 h | Match retrieval strategy to data shape for four cases |
| 24 | Ground with citations; evaluate RAG; decide RAG versus long context (3.6, 3.8, 4.2) | Grounding and citations · RAG evaluation · RAG versus long context | Write evaluation criteria for a RAG assistant | 2 h | State when long context beats RAG |
| 25 | Distinguish workflows from agents; apply workflow patterns (1.3, 1.5) | Workflows versus agents · Workflow patterns | Classify five fictional use cases as single call, workflow or agent | 1.5 h | Name five workflow patterns |
| 26 | Design the agent loop and multi-agent orchestration (1.4) | The agent loop · Multi-agent architecture | Lab 12: Agentic task execution | 2 h | Explain orchestrator-worker trade-offs |
| 27 | Apply human-in-the-loop; use managed agents and the Agent SDK; engineer reliability (5.3, 1.4) | Human in the loop · Managed Agents and Agent SDK · Agent reliability | Lab 13: Human in the loop | 2 h | Place approval gates in an agent that can send email and issue refunds |
| 28 | Revision milestone 4 | CCAR-P study guide D1 · CCAR-P study guide D3 | Half-length mock in learning mode across Domains 1 to 3 in the mock exam engine | 2 h | 75% or more overall |
Week 5 (days 29 to 35): AI security and governance¶
Focus: Domain 5.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 29 | Threat-model LLM systems; defend against direct and indirect prompt injection (5.1, 5.2) | LLM threat modelling · Prompt injection · Indirect prompt injection | Lab 14: Prompt injection testing | 2 h | Draw a trust-boundary diagram for a RAG agent |
| 30 | Prevent data disclosure, excessive agency and insecure output handling (5.1) | Data exfiltration and disclosure · Tool misuse and excessive agency · Insecure output handling | Lab 15: Sensitive data protection | 2 h | List controls for each of the three risks |
| 31 | Manage supply chain, agent identity and secrets (3.2, 5.1) | Supply-chain risks · Agent identity and permissions · Secrets management | Review a fictional deployment for secrets and identity weaknesses | 1.5 h | Explain short-lived credentials for agents |
| 32 | Logging, governance, human oversight and deployment controls (5.3, 3.4) | Logging and auditability · AI governance · Human oversight · Enterprise deployment controls · OWASP LLM mapping | Map your design from day 29 to the OWASP LLM risks | 2 h | Name the main OWASP LLM risks and a control for each |
| 33 | Ensure regulatory compliance; address bias, fairness and transparency (5.4, 5.5) | Data governance and compliance · CCAR-P study guide D5 | Write a data-handling summary for a fictional health-sector deployment | 1.5 h | Explain what changes when regulated data is in scope |
| 34 | Integrate security into an enterprise architecture (5.1, 1.2) | Lab 20: Secure enterprise architecture (README and design) | Lab 20: Secure enterprise architecture, first half | 2 h | Explain each control in the lab's reference design |
| 35 | Revision milestone 5 | Pages linked from missed questions | Domain 5 practice questions | 1.5 h | 80% or more on Domain 5 |
Week 6 (days 36 to 42): evaluation, testing and production operations¶
Focus: Domain 4 and Domain 3 operations.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 36 | Define evaluation metrics across accuracy, latency, cost, safety and security (4.1) | Why evaluate · Metrics | Define metrics for a fictional extraction service | 1.5 h | Name a metric in each of the five categories |
| 37 | Design eval datasets and grading methods (4.2) | Building eval datasets · Grading methods | Lab 16: Evaluation framework | 2 h | Explain code-based, model-based and human grading trade-offs |
| 38 | Run A/B tests; diagnose failures; run safety and red-team evals (4.3, 4.4) | A/B testing and iteration · Diagnosing failures · Safety and red-team evals | Diagnose three fictional failure reports as prompt, retrieval or model mismatch | 2 h | State a diagnosis order for a hallucination report |
| 39 | Prepare for production; build observability (3.4, 4.6) | Production readiness · Observability | Lab 19: Observability | 2 h | List what to log for every Claude call |
| 40 | Optimise cost and latency (4.5, 3.3) | Cost optimisation · Latency optimisation | Lab 17: Cost optimisation (revisit) | 1.5 h | Rank five cost levers for a given workload |
| 41 | Engineer reliability; plan capacity; respond to incidents; manage model upgrades (6.5, 7.3) | Reliability engineering · Rate limits and capacity · Incident response · Model upgrades and change management | Lab 18: Reliability engineering | 2 h | Outline a model-upgrade runbook |
| 42 | Revision milestone 6 | CCAR-P study guide D4 | Domain 4 practice questions; half-length mock across all domains | 2 h | 80% on Domain 4; 75% overall |
Week 7 (days 43 to 49): architecture, stakeholders and developer enablement¶
Focus: Domains 1, 6 and 7.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 43 | Translate business problems into solutions; follow a design process (1.1, 1.2, 6.1) | Solution design process · Architecture patterns | Run a structured discovery for a fictional client brief | 1.5 h | List your discovery questions from memory |
| 44 | Apply reference architectures and integration patterns (1.2, 1.3) | Reference architectures · Integration patterns | Adapt a reference architecture to the day-43 brief | 2 h | Justify the pattern you chose |
| 45 | Communicate decisions and trade-offs; write ADRs (6.2, 6.4) | Stakeholder communication · Architecture decision records · CCAR-P study guide D6 | Write one ADR for the day-44 design | 1.5 h | ADR sections from memory |
| 46 | Align to business value; manage the lifecycle and SLAs (1.6, 6.3, 6.5) | Business value and ROI · Lifecycle management | Estimate cost and value for the design; draft SLAs | 1.5 h | Explain how you would set expectations on accuracy |
| 47 | Configure Claude Code for teams (7.1) | What is Claude Code? · Setup and first session · CLAUDE.md and memory · Permissions and settings | Lab 10: Claude Code workflows | 2 h | Explain CLAUDE.md scope and permission modes |
| 48 | Use commands, Skills, subagents, hooks, MCP and CI with Claude Code (7.2, 7.3) | Commands, Skills, subagents · Hooks · MCP in Claude Code · Team rollout · Headless, CI and Agent SDK · CCAR-P study guide D7 | Design a team rollout plan with guardrails | 2 h | Name two hook use cases and one CI use case |
| 49 | Revision milestone 7: full mock 1 | None beforehand | Full-length timed CCAR-P mock (python 18-mock-examinations/run_mock_exam.py --track CCAR-P --mode timed) |
2.5 h | 70% or more in every domain; record your two weakest |
Week 8 (days 50 to 56): applied project and gap repair¶
Focus: All domains.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 50 | Scope a portfolio project: requirements, architecture, evaluation plan (1.1, 6.1) | Project 2: Knowledge-base Q&A assistant or Project 3: Security alert triage assistant | Write the project's scope and success metrics | 2 h | Success metrics cover quality, latency, cost and safety |
| 51 | Build the core path in mock mode (1.2, 3.5) | The chosen project's README and source | Implement and run the project's tests in mock mode | 2 h | Tests pass in mock mode |
| 52 | Add evaluation, guardrails and observability (4.1, 5.1, 4.6) | The chosen project's README | Add at least three eval cases and one guardrail; log usage | 2 h | You can explain every control to a reviewer |
| 53 | Practise architectural judgement on scenarios | CCAR-P decision scenarios · practical exercises | Answer every scenario in writing before reading the discussion | 2 h | Agree with the discussion on at least four in five |
| 54 | Repair your weakest domain | Common misconceptions · the study guide for that domain | 20 practice questions in that domain | 2 h | 80% on that domain |
| 55 | Repair your second-weakest domain | The study guide for that domain | 20 practice questions in that domain; retry missed with --weak |
2 h | 80% on that domain |
| 56 | Revision milestone 8: full mock 2 | None beforehand | Full-length timed mock using a different form | 2.5 h | 75% overall; no domain below 70% |
Days 57 to 60: final revision and readiness¶
Focus: All domains.
| Day | Objectives | Reading | Practical exercises | Time | Knowledge checkpoint |
|---|---|---|---|---|---|
| 57 | Consolidate key facts and trade-offs | Revision notes · CCAR-P one-pager · Model selection cheat sheet · AI security cheat sheet | Write a one-page summary of your own weak points | 1.5 h | Explain every item on your summary aloud |
| 58 | Fix remaining misses | Pages linked from missed questions | CCAR-P flashcards; retry missed questions (--weak) |
1.5 h | No repeated misses |
| 59 | Revision milestone 9: readiness | Readiness assessment | Third full mock (a different form); readiness assessment | 3 h | Above 80% overall with no domain below 70%, and "ready" |
| 60 | Prepare for exam day | Exam-day guide | Light review only; rest | 0.5 h | Logistics confirmed |
Revision milestones¶
| Day | Milestone | Pass bar |
|---|---|---|
| 7 | CCAO-F Domain 3 drill (15 questions) | 75% |
| 14 | Domain 2 questions | 80% |
| 21 | Domain 3 (tools and MCP) questions | 75% |
| 28 | Half-length mock, Domains 1 to 3 | 75% |
| 35 | Domain 5 questions | 80% |
| 42 | Domain 4 questions and half-length mock | 80% on Domain 4, 75% overall |
| 49 | Full timed mock 1 | 70% in every domain |
| 56 | Full timed mock 2 | 75% overall |
| 59 | Full timed mock 3 and readiness assessment | 80% overall, 70% per domain |
Practice scores are a study signal only; the questions are unofficial and do not predict an official scaled score.