---
title: "CCAR-F Hands-On Projects Before the Exam | Cred Farmer"
url: https://credfarmer.com/blog/ccar-f-hands-on-projects
description: "CCAR-F hands-on projects: the 4 exercises in Exam Guide v1.0 and Cred Farmer's 5 evening builds, one per domain of the Architect - Foundations exam."
last_updated: 2026-11-10
publisher: "Cred Farmer, an independent study platform, not affiliated with or endorsed by Anthropic or Pearson VUE"
---

# CCAR-F hands-on projects: what to build before the Architect - Foundations exam, domain by domain

CCAR-F hands-on projects: the guide’s four exercises, Cred Farmer’s five evening builds, why each should end in a written decision, and why none is part of the exam.

Published 10 November 2026Updated 10 November 20268 min readCovers all four exams

By the Cred Farmer editorial team.Checked against the official sources on 10 November 2026.

Short answer

CCAR-F hands-on projects start in the CCAR-F Exam Guide v1.0: its section 8 sets four preparation exercises that together name all five domains. Cred Farmer’s suggestion cuts the same ground into five builds, one per domain, an evening each. Each of those builds ends in one decision a task statement describes, such as a hook or a prompt for a rule that must hold. You take that decision once before an exam item may ask you to choose it. No build is submitted or marked.

**What the exam codes mean**

**CCAO-F**

Claude Certified Associate - Foundations

**CCDV-F**

Claude Certified Developer - Foundations

**CCAR-F**

Claude Certified Architect - Foundations

**CCAR-P**

Claude Certified Architect - Professional

Checked against the official sources on 10 November 2026: the [CCAR-F Exam Guide v1.0](https://everpath-course-content.s3-accelerate.amazonaws.com/instructor%2F6nizmqk8tpzpfjvt6qmmav7rh%2Fpublic%2F1783542750%2FClaude+Certified+Architect+%E2%80%93+Foundations+Exam+Guide.pdf) (effective July 2026) and Anthropic’s [certification FAQ](https://anthropic-partners.skilljar.com/page/faq-certifications). The sources also include Partner Academy’s [certifications page](https://anthropic-partners.skilljar.com/page/partner-certifications), [CCAR-F registration page](https://anthropic-partners.skilljar.com/claude-certified-architect-foundations-certification) and [Architect path](https://anthropic-partners.skilljar.com/path/claude-architect), and the [Claude Agent SDK overview](https://platform.claude.com/docs/en/agent-sdk/overview). Cred Farmer publishes CCAR-F practice papers, so it has a stake in how you prepare.

## What hands-on projects should I build before the Architect exam?

For the [Claude Certified Architect - Foundations (CCAR-F)](https://credfarmer.com/certifications/claude-certified-architect-foundations) exam, the guide itself names what to build. Section 7 opens with “Build an agent with the Claude Agent SDK” and “Configure Claude Code for a real project”, and section 8 sets four exercises. Cred Farmer’s suggestion adds five evening builds, one per domain.

| Domain | Weight | Guide exercises naming it | Cred Farmer’s build |
| --- | --- | --- | --- |
| 1 [Agentic Architecture & Orchestration](https://credfarmer.com/blog/agentic-architecture-on-ccar-f) | 27% | 1 and 4 | A loop with a gate |
| 2 [Tool Design & MCP Integration](https://credfarmer.com/blog/tool-design-and-mcp-on-ccar-f) | 18% | 1, 2 and 4 | Two tools that look alike |
| 3 [Claude Code Configuration & Workflows](https://credfarmer.com/blog/claude-code-configuration-and-workflows-on-ccar-f) | 20% | 2 | One rule, four places |
| 4 [Prompt Engineering & Structured Output](https://credfarmer.com/blog/prompt-engineering-and-structured-output-on-ccar-f) | 20% | 3 | An extractor allowed to return null |
| 5 [Context Management & Reliability](https://credfarmer.com/blog/context-management-and-reliability-on-ccar-f) | 15% | 1, 3 and 4 | A conversation that keeps its facts |

Weights are from the guide’s section 4 and exercise numbers from its section 8. Each domain name links to Cred Farmer’s lesson on that domain.

## CCAR-F hands-on projects start in the guide’s section 8

The CCAR-F Exam Guide v1.0 is the only one of the four v1.0 guides with a section of preparation exercises. Its section 8 sets four, each with its steps and a “Domains reinforced” line, and opens: “Complete these hands-on exercises to build practical familiarity with the topics covered on the exam.”

Exercises 1 to 4 are “Build a Multi-Tool Agent with Escalation Logic”, “Configure Claude Code for a Team Development Workflow”, “Build a Structured Data Extraction Pipeline” and “Design and Debug a Multi-Agent Research Pipeline”. Anthropic’s FAQ calls the exam guide “the authoritative source for exam scope”.

**Figure:** A grid of the CCAR-F guide's four preparation exercises against its five domains. Exercise 1 names Domains 1, 2 and 5. Exercise 2 names Domains 2 and 3. Exercise 3 names Domains 4 and 5. Exercise 4 names Domains 1, 2 and 5. The weights are 27, 18, 20, 20 and 15 percent

- Which of the five CCAR-F domains each of the guide’s four exercises names
- D1 27% | D2 18% | D3 20% | D4 20% | D5 15%
- Exercise 1 Multi-tool agent: Yes | Yes | blank | blank | Yes
- Exercise 2 Claude Code set-up: blank | Yes | Yes | blank | blank
- Exercise 3 Extraction pipeline: blank | blank | blank | Yes | Yes
- Exercise 4 Research pipeline: Yes | Yes | blank | blank | Yes

Yes means that exercise’s Domains reinforced line names the domain. Weights are from section 4.

By Cred Farmer’s count of those lines, Domain 1 is named by two exercises, Domains 2 and 5 by three each, and Domains 3 and 4 by one each, though each weighs 20%. Skip Exercise 2 and no exercise you did names Domain 3. Skip Exercise 3 and the same goes for Domain 4. That gap is why Cred Farmer’s builds go one per domain.

## Cred Farmer’s five evening builds, one domain each

Cred Farmer’s five CCAR-F builds give each domain one evening, time-boxed to about three hours, and each names the task statement it targets. A build is over once one decision is taken and its reasons written down, finished code or not. All five are Cred Farmer’s suggestion: do them first, then the guide’s fuller exercises.

1

### Domain 1: a loop with a gate

Give an agent a look-up tool and an action tool, and end its loop on `stop_reason`, not a turn cap. Make the action wait for a verified look-up, first by a prompt line, then by a hook. Decide which you would ship. Targets 1.5: “Apply Agent SDK hooks for tool call interception and data normalization”.

2

### Domain 2: two tools that look alike

Write two MCP tools whose jobs overlap, with one-line descriptions, and send ambiguous requests. Add inputs, an example and a boundary to each description, resend, and decide whether that alone fixed the routing. Targets 2.1: “Design effective tool interfaces with clear descriptions and boundaries”.

3

### Domain 3: one rule, four places

In a repository you own, put one instruction each in your user CLAUDE.md, the project CLAUDE.md, a `.claude/rules/` file with a paths glob, and a skill set to `context: fork`. From a fresh session, run `/memory` and edit a matching file. Decide where each belongs. Targets 3.3: “Apply path-specific rules for conditional convention loading”.

4

### Domain 4: an extractor allowed to return null

Give an extraction tool a JSON schema with required fields, nullable fields and an enum carrying “other” and a detail field. Feed it documents with facts missing, and check for null, not invented values. Add a retry carrying the validation error, and decide when a retry can work. Targets 4.4: “Implement validation, retry, and feedback loops for extraction quality”.

5

### Domain 5: a conversation that keeps its facts

Run a long support-style conversation through the API and summarise its older turns. Keep amounts, dates and order numbers in a facts block sent with every request, and decide what must stay out of the summary. Targets 5.1: “Manage conversation context to preserve critical information across long interactions”.

Cred Farmer’s advice is to run every build locally and skip deployment. The guide’s section 17 puts deploying or hosting MCP servers among 16 topics, by Cred Farmer’s count, that “will not appear on the exam”. See [the Claude exam out of scope list](https://credfarmer.com/blog/claude-exam-out-of-scope), and [a Claude practice environment](https://credfarmer.com/blog/claude-practice-environment) for access and billing.

## Why Cred Farmer’s builds end in a written decision

Cred Farmer’s builds end in a written decision because the CCAR-F guide frames the credential around decisions. Section 1 says the certification “validates that practitioners can make informed decisions about tradeoffs when implementing real-world solutions with Claude”. Section 6 says exam items “are written against these objectives”, meaning the task statements.

### Cred Farmer’s finish line for a build

A build is done when its log names the choice, the task statement and why each alternative falls short. Finished code is optional.

The guide’s 12 sample questions model that last part, because each answer says where the other options fail. Cred Farmer’s reason for building is that you will have taken each trade-off once yourself, in code, before an item may offer it. No figure measures what that does for a score, and Cred Farmer promises none.

The [Claude certification study guide](https://credfarmer.com/blog/how-to-study-for-the-claude-certification) has the review loop that follows a build.

## Do five builds stand in for 6+ months of experience?

No: five evening builds do not replace the experience the CCAR-F guide describes. Its candidate “typically has 6+ months of practical experience building with Claude APIs, Agent SDK, Claude Code, and MCP”. Five evenings are not six months. Cred Farmer offers the builds as one pass over each domain, not as a stand-in.

That line describes the guide’s ideal candidate, “a solution architect who designs and implements production applications with Claude”. The v1.0 CCAO-F, CCDV-F and CCAR-P guides call their experience lines “recommended, not required”, and the CCAR-F guide prints no such line. Partner Academy’s Architect path, which it calls a start toward the exam, says “No Claude experience is assumed”.

## A CCAR-F build is preparation, not part of the exam

Nothing you build is part of the CCAR-F exam. It sets 120 minutes for 60 multiple-choice and multiple-response items, and section 10 of the guide bases the result “on your total scaled score”. No page Cred Farmer read on 10 November 2026 offers a review of a project for this exam.

Cred Farmer’s suggestion is to do each build after studying its domain and before a timed practice paper. The [two-week CCAR-F study plan](https://credfarmer.com/blog/two-week-study-plan-ccar-f) sets the days, the [Claude certification study plan](https://credfarmer.com/guides/claude-certification-study-plan) covers longer runs, and the [CCAR-F exam scenarios](https://credfarmer.com/blog/ccar-f-exam-scenarios) show the six contexts, four per sitting.

Anthropic’s notice of updated content from 7 December 2026 names the [Claude Certified Associate - Foundations (CCAO-F)](https://credfarmer.com/certifications/claude-certified-associate-foundations), [Claude Certified Developer - Foundations (CCDV-F)](https://credfarmer.com/certifications/claude-certified-developer-foundations) and [Claude Certified Architect - Professional (CCAR-P)](https://credfarmer.com/certifications/claude-certified-architect-professional) exams. That notice omits CCAR-F, whose guide read on 10 November 2026 was still v1.0: see [Claude certification changes](https://credfarmer.com/blog/claude-certification-changes).

[Claude certification prep courses](https://credfarmer.com/blog/claude-certification-prep-courses) covers the free courses the CCAR-F page lists. For a firm running practice projects on its own material, see [internal Claude projects](https://credfarmer.com/blog/internal-claude-projects).

## Key takeaways

- **The CCAR-F Exam Guide v1.0 is the only v1.0 guide with a section of preparation exercises.** Its section 8 sets four, and together they name all five domains.
- **By Cred Farmer’s count, Domains 3 and 4 of CCAR-F are each named by one guide exercise.** Both weigh 20%, so skipping Exercise 2 or 3 leaves one of them with no exercise.
- **Cred Farmer suggests five evening builds for CCAR-F, one per domain.** Each ends in a written decision tied to a task statement, finished code or not.
- **The CCAR-F guide’s candidate “typically has 6+ months of practical experience”.** Five evenings do not replace that, and the guide prints no line calling it a requirement.
- **Nothing you build is part of the CCAR-F exam.** The result comes from your total scaled score on 60 multiple-choice and multiple-response items.

## Find the domain to build first

The free CCAR-F mock is 20 questions in 40 minutes, no account needed, with a score for each of the five domains. A free sitting saves no result, so note the weakest domain it names and start with that build. The [practice exams guide](https://credfarmer.com/guides/claude-certification-practice-exams) lists the other papers.

[Try the free CCAR-F mock: 20 questions in 40 minutes](https://credfarmer.com/practice/ccar-f)

## Frequently asked

**Are Claude architect exam projects required to sit CCAR-F?**

No. Anthropic’s FAQ sets the conditions for sitting a Claude exam: an employer in the Claude Partner Network, a partner email on a recognised company domain, and an age of at least 18. None is a project. The CCAR-F guide says it “recommends preparation strategies”, and its exercises are among them.

**What to build before CCAR-F if I only have one evening?**

Cred Farmer’s suggestion is the build for the domain your practice paper scores lowest. With no paper sat yet, take Domain 1, Agentic Architecture & Orchestration, the heaviest at 27% of the exam. One evening covers one domain of five, so it is a start, not coverage.

**Does CCAR-F practical preparation need an API key?**

Four of Cred Farmer’s five builds do: they call the Claude API or run on the Claude Agent SDK, whose overview says to “set your API key”. The fifth runs in Claude Code. The CCAR-F guide rules “Claude API authentication, billing, or account management” out of the exam itself.

**Will Anthropic review or grade my CCAR-F project?**

No page Cred Farmer read on 10 November 2026 offers that. The CCAR-F exam is 60 multiple-choice and multiple-response items, and its guide bases the result on your total scaled score. Partner Academy’s Architect path says that, after the separate Claude Code partner badge’s courses, “a capstone and an assessment follow”.

**How long should each CCAR-F build take?**

Cred Farmer time-boxes each build to one evening, about three hours, and stops once the decision is written down. The CCAR-F guide gives no time for its own four exercises. If a build runs over, Cred Farmer’s advice is to cut its scope and keep its decision.

## Related posts

### [CCAR-P vs CCAR-F: what actually changes at the Professional level](https://credfarmer.com/blog/ccar-p-after-ccar-f)

CCAR-P vs CCAR-F using the current blueprints: five domains versus seven, what transfers, what Professional adds, and why difficulty is personal.

### [The week before the Claude exam: what to revise, what to stop, and what to check each day](https://credfarmer.com/blog/week-before-the-claude-exam)

A 7-day plan for the week before the Claude exam, by Cred Farmer: the ID and 48-hour reschedule checks on day 1, 2 timed papers, and nothing new after day 5.

### [A Claude certification study guide you can follow in 30 days](https://credfarmer.com/blog/how-to-study-for-the-claude-certification)

A Claude certification study guide and 30-day plan: four passes a day, time sized to blueprint weights, and practice evidence used to plan the next review.

Not affiliated with, or endorsed by, Anthropic or Pearson VUE. Details are summarised from publicly published program information and can change. Always confirm against the official exam guide before booking.