Claude Certified Architect Exam: Cost, Blueprint and Who Can Register (2026)

Claude Certified Architect – Foundations (exam code CCAR-F) is Anthropic’s exam for people who build production systems with Claude Code, the Claude Agent SDK, the Claude API and the Model Context Protocol (MCP). According to Anthropic’s exam guide (version 1.0, effective July 2026), it has 60 multiple-choice and multiple-response items in 120 minutes, is scored from 100 to 1,000 with 720 to pass, costs $125 per attempt and stays valid for 12 months. Pearson VUE delivers it online or at a test center.

One rule decides whether any of that applies to you. Anthropic’s certification FAQ says only people at Claude Partner Network organizations can register, with a partner email address on a recognized company domain. If your employer is not a partner, you cannot book the exam today.

The Professional tier (CCAR-P) is a separate exam. Its exam guide lists 63 items and a $175 fee, and the FAQ says there is no requirement to pass Foundations first.

The gate is your employer, not your experience. Anthropic’s FAQ answers “Do I need to work at a partner organization to take an exam?” with:

“Yes. Certification is available to people at Claude Partner Network organizations. Registration requires a partner email address on a recognized company domain — personal email addresses will not work.”

The same FAQ adds the details that trip people up at registration:

• If the site says you aren’t signed in with a partner company employee email, first check that you used your company address rather than a personal one. If you did, the domain is probably not on your company’s partner record, and adding it “takes 7 to 10 days to take effect”, so ask well before you plan to sit.
• Subsidiary and regional domains do not credit to your company’s partner account unless they are rolled up to the parent domain.
• You must be at least 18, and the name on your government ID must match your registration exactly.
• Candidates with IDs from Belarus, Cuba, North Korea, Russia, Syria or restricted regions of Ukraine cannot test online and must use a Pearson test center. Since September 8, 2026, Pearson has suspended delivery for residents of Iran.
• A credential you earn can move with you: the FAQ says it can transfer if you change email addresses or organizations, and recommends adding a personal email to your Credly profile. The Terms still let Anthropic revoke a certification in some cases, including when your partner is no longer in good standing.

On experience, the Foundations guide says “The candidate typically has 6+ months of practical experience building with Claude APIs, Agent SDK, Claude Code, and MCP” and names no required course. Unlike the other three guides, it does not say whether that experience is required, and the FAQ lists no experience check at registration.

Everything below is from Anthropic’s own documents, read on September 26, 2026. The source column names the document for each row.

The exam guide splits the exam into five weighted domains and breaks each into task statements, the “Knowledge of” and “Skills in” lists that exam items are written against. The weights are the approximate share of scored items per domain.

Two things follow from the arithmetic. First, there are 30 task statements for 60 items, an average of 2 items’ worth each. The guide sets weights only per domain, so that is an average rather than a promise, but it leaves little room for a task statement to be a footnote. Second, the smallest domain, Context Management & Reliability at 15%, still carries about 9 items, and it is a primary domain in more scenarios than any other, as the next section shows.

What will not be on it

The guide’s appendix is specific about scope. It lists 16 topics that “will not appear on the exam”, which is as useful for planning as the blueprint itself:

• Fine-tuning Claude models or training custom models
• Claude API authentication, billing, or account management
• Detailed implementation of specific programming languages or frameworks (beyond what's needed for tool and schema configuration)
• Deploying or hosting MCP servers (infrastructure, networking, container orchestration)
• Claude's internal architecture, training process, or model weights
• Constitutional AI, RLHF, or safety training methodologies
• Embedding models or vector database implementation details
• Computer use (browser automation, desktop interaction)
• Vision/image analysis capabilities
• Streaming API implementation or server-sent events
• Rate limiting, quotas, or API pricing calculations
• OAuth, API key rotation, or authentication protocol details
• Specific cloud provider configurations (AWS, GCP, Azure)
• Performance benchmarking or model comparison metrics
• Prompt caching implementation details (beyond knowing it exists)
• Token counting algorithms or tokenization specifics

Architect – Foundations is the only one of the four Claude exams whose guide documents a scenario bank. Every question hangs off a realistic production scenario, and you do not get all six:

“During the exam, 4 scenarios are presented and picked at random from the full set of the 6 scenarios below.”

The six are: Customer Support Resolution Agent; Code Generation with Claude Code; Multi-Agent Research System; Developer Productivity with Claude; Claude Code for Continuous Integration; Structured Data Extraction. With 4 of 6 drawn, there are 15 possible combinations. Here is what that means if you decide to study only some of them:

Skipping one scenario leaves a 67% chance it is on your exam. Skipping two raises that to 93%, because only one of the 15 combinations avoids both. Skip three and it is certain.

The guide also lists each scenario’s primary domains. Context Management & Reliability is primary in four of the six, so any draw of four includes at least two scenarios built around it. Prompt Engineering & Structured Output is primary in only two (Continuous Integration and Structured Data Extraction), and 7% of draws, 1 in 15, include neither.

Anthropic now runs four exams, each with its own v1.0 guide. They share a format (120 minutes, pass at 720, 12-month validity) and differ in almost everything else. The table sets them side by side; the arithmetic rows are ours.

Reading the four blueprints against each other turns up differences a summary would hide:

• The Developer exam gives Claude Code far less room. The domain named for Claude Code is 20% of Architect – Foundations and 3.1% of Developer – Foundations, whose biggest domain is Applications and Integration at 33.1%. Some Claude Code topics sit elsewhere in the Developer guide, such as CLAUDE.md and settings.json under Configuration Management (4.1%).
• Evaluation moves up a tier. “Eval, Testing, and Debugging” is 2.6% of the Developer exam; “Evaluation, Testing & Optimization” is 16% of Architect – Professional.
• Professional is partly not a build exam. Governance, Safety & Risk Management and Stakeholder Communication & Lifecycle Management are 14% each, 28% of the blueprint together.
• Only the Architect – Foundations samples come from a practice test. Its guide prints 12 sample questions “drawn from the practice test”; the other three guides print 3 each and say “They are not drawn from the live item bank.”
• The Associate is the cheapest and the only one that does not count toward a company’s Claude Partner Network eligibility, which may matter if your employer is paying.

“Foundations is the natural place to start and there’s no formal prerequisite, so you can take Professional without holding Foundations. Foundations does not convert or upgrade to Professional automatically.”

Which one fits you

Going by each guide’s own audience description:

• You build agents with the Agent SDK, configure Claude Code for a team and design MCP tools: Claude Certified Architect – Foundations.
• You ship applications through the API and SDKs in Python or TypeScript: Claude Certified Developer – Foundations.
• You own architecture, governance and stakeholder decisions with 3+ years in systems architecture: Claude Certified Architect – Professional.
• You use Claude for business work and do not code: Claude Certified Associate – Foundations. Our free Claude Certified Associate practice test follows that guide’s seven domains.

The list price is $125, and every attempt is charged in full. Your employer’s partner tier sets the discount:

“Registered-tier partners pay full price. Select, Preferred, and Global Premier partners receive 50% off, applied automatically at checkout. Through December 31, 2026, Global Premier partners receive a 100% discount on all certification exams; after that date, the standard 50% discount applies.”

The FAQ adds: “Your partner tier discount applies to retakes.” So the running total for up to four attempts in a year looks like this:

At list price, four attempts at Architect – Foundations cost $500; at a 50% tier, $250. Four attempts at Professional would cost $700 at list.

Booked as early as the waits allow, a fourth attempt falls on day 134. After that, the rolling 12-month limit applies until your first attempt is a year old. Three other costs are easy to miss:

• A no-show, a late arrival or a name that does not match your ID forfeits the fee.
• Cancelling in Pearson VUE does not refund you. The FAQ says to email certifications support to request a refund.
• Renewal is free only if it is on time; a lapsed credential means the full exam at the full fee. The guide adds: “If exam content changes significantly, Anthropic may require holders to retake the full exam to recertify rather than complete the renewal assessment.” The FAQ says full renewal details “will be shared before the first certifications come up for renewal.”

Anything written about this exam before June 30, 2026 describes rules that have since changed. That day the program moved to Pearson VUE, and Anthropic’s FAQ ties four changes to the move:

• Price: Architect – Foundations went from $99 to $125. Registrations completed before June 30 were charged $99.
• Validity: from 6 months to 12. Credentials earned under the 6-month rule were extended automatically to 12 months from the date earned.
• Practice exam: retired in the move. Score reports from exams taken on the old ProctorFree platform are no longer available.
• Delivery and badges: Pearson VUE for the exam, Credly for the badge.
• A clean slate for earlier failures: anyone who attempted Architect – Foundations before June 30 and did not pass has that attempt cleared and can retake on Pearson “with no waiting period”.

“The practice exam available on the previous platform was retired in the move to Pearson. The exam guide includes sample questions that show the format and style of what’s on the exam.”

One more trace of the earlier program: some third-party pages still call the exam “CCA-F”. The current guide uses CCAR-F, and the Pearson VUE page lists all four codes: CCAO-F, CCAR-F, CCAR-P and CCDV-F.

Our keyword research pulls monthly US search volumes for the exams we write about. For the Claude credentials the pattern lines up with Anthropic’s own document history: searches for the Architect name start in February–March 2026, when the Foundations guide’s first draft is dated, and the other three names take off in July 2026, the month their guides were first published.

In August 2026 the Architect name drew 22,200 US Google searches against 1,600 for the Associate, about 14 to 1, after peaking at 40,500 in March. Bing shows the same order: 1,600 against 140. With that much interest, many third-party guides compete for these searches, which is why we checked what they say.

On September 26, 2026 we ran five common US searches for this exam, collected the 15 third-party pages they returned (prep sellers, study sites, consultancies, blogs and a newsletter) and compared each with the official documents. 2 refused our reader, so 13 were checked.

The headline numbers have caught up. None of the 13 still presents $99 or 6-month validity as current; 8 give the $125 fee and 10 give 12-month validity. The gaps are in the rules that decide whether and when you can sit:

• 5 of 13 never say that only Claude Partner Network employees can register.
• 2 still send you to the official practice exam that was retired in the June 30 move; 2 say it is no longer available.
• 1 gives a different retake rule: a 14-day wait between attempts, capped at three attempts in the first 12 months. 5 give the full 14/30/90-day rule with four attempts.
• 3 give a change deadline, all of them using the guide’s 24 hours. None mentions the FAQ’s stricter 48.

If you are reading a guide to this exam, including this one, check those four rules against the FAQ before you register. They are the ones that cost money when they are wrong.

With the practice exam retired, the Anthropic material aimed at this exam is the prep courses linked from the certification page, the 12 sample questions with explanations in the exam guide, and four hands-on exercises in the same guide. Each exercise names the domains it reinforces:

The coverage is uneven. Tool Design & MCP and Context Management each get three of the four exercises; Claude Code Configuration and Prompt Engineering, each 20% of the exam, get one apiece. If either of those is your weak area, budget your own hands-on time for it: the guide’s §7 suggests configuring Claude Code on a real project with a CLAUDE.md hierarchy, path-specific rules, custom skills and an MCP server, and building an extraction pipeline with JSON schemas and validation-retry loops.

Documents. We downloaded the four exam guides, the Exam Policy and the Terms and Conditions, and read the certification FAQ and program pages, on September 26, 2026. The guides say they are “subject to change without notice”, so the Sources list below gives each PDF’s SHA-256 hash: if Anthropic re-issues a file under the same link, the hash will differ.

Arithmetic. Items per domain are weight × items. Scenario odds count the 15 equally likely ways to draw 4 of 6. Retake dates add the published waits to day 0. Costs multiply the fee, less the partner discount, by the number of attempts. Each calculation is covered by automated tests that pin the printed values.

Search data. Monthly US volumes pulled on September 23, 2026 through DataForSEO’s Google and Bing keyword data. They are estimates, rounded by the providers.

Guide check. Done on September 26, 2026 as described above. Every page we marked as out of date or wrong was re-read in full before it was counted.

What we did not do. We have not sat a Claude certification exam, and nothing here describes live exam content, which the exam’s confidentiality agreement protects. We did not test checkout discounts. Anthropic publishes no pass rate and does not say whether any of the 60 items are unscored.

Written by the Every Exam Prep editorial team and drafted with AI from the documents listed below, as our editorial policy describes. Found an error or a changed rule? Email contact@everyexamprep.com.