Skip to content

Claude Certified Architect – Foundations (CCAR-F) Preparation Track

Unofficial study track. This repository is not affiliated with or endorsed by Anthropic. The official CCAR-F Exam Guide v1.0 is the source of truth. Every exam fact on this page comes from registry.yaml (verified 2026-10-24), exam-facts.md and the official certifications page.

What the credential is for

CCAR-F is the foundations-level architecture credential in Anthropic's Claude Certification Program. It signals that you can build and configure production systems with the four technologies architects reach for most: the Claude Agent SDK (agent loops, subagents, hooks, sessions), Claude Code (CLAUDE.md, rules, skills, plan mode, CI), the Model Context Protocol (tool and resource design, structured errors) and the Claude API (tool use, structured output, the Message Batches API).

"Foundations" does not mean "easy". The exam is practical: it asks what you would configure, build or change in a specific production situation, and the wrong answers are usually things that work in a demo but fail at scale, such as relying on a prompt instruction where a hook is needed, or giving every agent every tool.

Who it is for

The registry describes the target candidate as a solution architect who designs and implements production applications with Claude, with hands-on experience of the Agent SDK, Claude Code, MCP tool design and structured-output prompting. In practice this track also suits:

  • Senior developers and tech leads who own a team's Claude Code setup or an internal agent.
  • Platform and DevSecOps engineers who wire Claude Code into CI/CD and govern MCP servers.
  • Security engineers who review agent designs and want to know where enforcement really lives.

How CCAR-F differs from CCAR-P

CCAR-F (this track) CCAR-P
Focus Building and configuring: SDK, Claude Code, MCP, structured output Designing and governing whole solutions across an enterprise
Typical question "Which configuration, hook, schema or flag fixes this?" "Which architecture best fits these constraints and stakeholders?"
Format Scenario-based: 4 scenarios drawn from a published bank of 6 Independent items across seven domains

Many candidates take CCAR-F first; the hands-on knowledge transfers directly to CCAR-P's Domain 3 (Integration) and Domain 7 (Developer Productivity).

From the registry: no mandatory prerequisites, but 6+ months of practical experience building with the Claude API, Agent SDK, Claude Code and MCP is recommended. If you have used Claude Code but never written an agent loop or an MCP server, budget extra time for labs 03, 08, 09, 10 and 12 in 19-hands-on-labs.

Format and logistics

Fact Value (from registry.yaml)
Exam code CCAR-F
Guide version 1.0 (effective July 2026)
Items 60
Item types Multiple-choice and multiple-response (each item says how many to select)
Structure 4 scenarios drawn at random from a published bank of 6
Time 120 minutes
Passing standard Scaled score 720 on a 100–1,000 scale
Fee USD 125
Validity 12 months
Result Pass/fail, scaled score and percent-correct by domain
Delivery Pearson VUE, online proctored (OnVUE) or test centre

See exam-facts.md for all tracks side by side and exam-day.md for check-in and pacing.

The scenario-based format: 4 of 6

This is the feature that most changes how you prepare.

  • The exam guide publishes six scenarios, each a short description of a production context. The registry lists them with their primary domains:
Scenario Primary domains
Customer Support Resolution Agent 1, 2, 5
Code Generation with Claude Code 3, 5
Multi-Agent Research System 1, 2, 5
Developer Productivity with Claude 2, 3, 1
Claude Code for Continuous Integration 3, 4
Structured Data Extraction 4, 5
  • On exam day, four of the six are chosen at random. Each scenario frames a group of questions, so you read the context once and then answer several items about it.
  • You cannot predict which four you will get. Every scenario has a roughly one-in-three chance of being left out, so skipping one in your preparation is a gamble you will lose about two times in three.
  • Domain weights still apply across the whole exam. A scenario's "primary domains" tell you where most of its questions come from, not that other domains are excluded.

What this means for study. Prepare every scenario as a mini case study: know the reference architecture, the decisions an architect makes in it and the traps. scenarios.md does this for all six, with original content (no exam items). Then practise reading a scenario once and holding its constraints in mind across several questions, which is what the practice questions and mock examinations for this track train.

Registration and eligibility

  • Eligibility: certification exams are open to organisations in the Claude Partner Network. The free Claude Academy courses are open to everyone but award completion certificates, not certifications.
  • Registration path: enrol through the Anthropic Partner Academy (Skilljar) certification page, then schedule with Pearson VUE. Your Pearson profile name must exactly match your government-issued ID.
  • Rescheduling: the v1.0 guide says changes are allowed up to 24 hours before the appointment, but the Partner Academy FAQ and Pearson VUE page (checked 2026-10-24) require at least 48 hours. The registry treats 48 hours as the effective deadline; later changes forfeit the fee.

Retake and renewal policy

  • Retakes (registry): wait 14 days after the first failed attempt, 30 days after the second and 90 days after the third, with at most four attempts per exam in a rolling 12 months. Each attempt is charged.
  • Validity: 12 months from the award date.
  • Renewal (exam guide v1.0, section 15, checked 2026-10-24): renewing before expiry is free. You review what has changed and pass a non-proctored check on the Anthropic Partner Academy. A lapsed credential means sitting and paying for the full exam again, and Anthropic may require the full exam if content changes significantly.

Upcoming change: 2026-11-17 guide update

Pearson VUE states that updated exam guides will be published on 2026-11-17 and names CCAO-F, CCDV-F and CCAR-P as receiving updated content from 2026-12-07. CCAR-F is not named in that notice.

Unverified: whether the CCAR-F guide will also change on 2026-11-17. Check the certifications page before you book, and compare any new guide's objectives with registry.yaml.

Product facts move faster than the exam guide

The v1.0 guide was written against the products as they stood in mid-2026. A few details have since changed in the official docs, and this track teaches both the concept the exam tests and the current behaviour (as of 2026-10-24):

Exam-guide concept Current behaviour to know
Subagents are spawned with the "Task" tool Use the Agent tool, called Task in the v1.0 exam guide (the system:init tool list still shows Task). List Agent in allowedTools so delegation runs without a prompt
allowedTools for a coordinator allowedTools auto-approves the listed tools; it does not restrict the tool set. Restrict with AgentDefinition.tools, disallowedTools or deny rules
Forcing a tool with tool_choice any or {"type": "tool"} Returns a 400 on Claude Opus 5.5, Sonnet 5.5, Fable 5.1 and Mythos 5.1; Haiku 5.5 still accepts it. Use auto plus an explicit instruction, strict: true tools, or structured outputs
Skills' allowed-tools frontmatter Pre-approves listed tools; it does not remove others. Use disallowed-tools to restrict
Slash commands in .claude/commands/ Still work, but are now merged into skills; a skill with the same name wins

The domain study guides explain each one where it arises.

Domains

# Domain Weight Study guide Key modules
1 Agentic Architecture & Orchestration 27% domain-1.md 10 Agentic AI, 09 Hooks
2 Tool Design & MCP Integration 18% domain-2.md 07 Tool use, 08 MCP
3 Claude Code Configuration & Workflows 20% domain-3.md 09 Claude Code
4 Prompt Engineering & Structured Output 20% domain-4.md 03 Prompt engineering, 06 Batches
5 Context Management & Reliability 15% domain-5.md 04 Context engineering, 10 Agent reliability

Domain 1 alone is more than a quarter of the exam, and it appears as a primary domain in four of the six scenarios. Domains 3 and 4 together are another 40%.

The order below follows dependencies, not domain numbers.

  1. Baseline (week 1). Take the readiness assessment cold and skim revision-notes.md. Read scenarios.md once to see the six contexts.
  2. Tool use and structured output (Domain 4, objectives 4.3–4.4; Domain 2, 2.1–2.3). Everything agentic rests on a correct tool loop and well-designed tools. Do labs 03 and 08.
  3. Agent loop and orchestration (Domain 1). Loop control on stop_reason, coordinator and subagents, hooks, decomposition, sessions. Do lab 12 and lab 13.
  4. MCP (Domain 2, 2.2 and 2.4). Build a small server with structured errors and connect it to Claude Code. Do lab 09.
  5. Claude Code (Domain 3). CLAUDE.md hierarchy, .claude/rules/, skills, plan mode, headless CI. Do lab 10 end to end on a real repository.
  6. Prompting for precision (rest of Domain 4). Explicit criteria, few-shot examples, batches, multi-pass review. Do labs 02, 16 and 17.
  7. Context and reliability (Domain 5). Case-facts blocks, escalation, error propagation, large-codebase exploration, confidence calibration, provenance. Do labs 05 and 18.
  8. Consolidate by scenario. Work through scenarios.md, decision-scenarios.md, practical-exercises.md, common-misconceptions.md, the CCAR-F flashcards, the practice questions and the mock examinations.
  9. Final week. Re-take the readiness assessment, remediate weak objectives, read exam-day.md.

For a calendar-based plan, follow Weeks 1 to 4 of the 60-day plan and add Module 09 (Claude Code); the roadmaps guide explains why.

Track contents

File Purpose
objective-map.yaml Every CCAR-F objective mapped to modules and labs
coverage-matrix.md Generated matrix of objectives, modules, labs and practice questions
scenarios.md Study guide for each of the six published exam scenarios
study-guide/ Five domain guides
practical-exercises.md Hands-on build exercises (agent loop, hooks, MCP server, Claude Code setup, CI review, extraction pipeline)
decision-scenarios.md Original configuration case studies with options analysis
revision-notes.md Condensed notes by domain
common-misconceptions.md Misconceptions and corrections
CCAR-F one-pager Single-page cheat sheet
exam-day.md Logistics, rules and pacing for a scenario-based exam
readiness-assessment.md Per-objective self-assessment and remediation

Official resources