Free sample exam
Claude Certified Architect – Foundations
A free sample of the questions on Anthropic's certification for people who build with Claude, scored as you go, with the reason for every answer.
Tecyfy is independent of Anthropic. These are our own practice questions, written to the topics in the official exam guide.
The real exam
The exam tests whether you can choose the most reliable, proportionate design for a production system built with Claude: agent loops, tools and MCP, Claude Code, prompts and structured output, and keeping long tasks reliable. Every question sits inside a scenario.
Only people who work at an organisation in the Claude Partner Network can sit it: you register with a company email on a domain in your organisation's partner record. Anyone can use the sample below.
- Questions
- 60, multiple choice and multiple response
- Time
- 120 minutes
- Pass mark
- 720 on a scale of 100 to 1,000
- Scenarios
- 4 of 6 per sitting, chosen at random
- Fee
- $125
- Valid for
- 12 months, renewed free with an online check
- Who can sit it
- Staff of Claude Partner Network organisations
What it covers, by share of the questions
- Agentic architecture and orchestration27%
- Tool design and MCP integration18%
- Claude Code configuration and workflows20%
- Prompt engineering and structured output20%
- Context management and reliability15%
The sample
10 questions in the style of the exam. Pick an answer to see why it is right or wrong. Your best score stays in this browser.
Question 1 of 10
Question 1
Developer productivity. Your agent's own loop ends a run as soon as a response contains text. Engineers say it often stops right after writing "Let me check the imports next", without checking anything. What should decide whether the loop goes on?
Not this time. One response can hold text and tool calls together, so text is no sign the work is over.
stop_reasonsays it:tool_usemeans Claude wants a tool run and the result sent back,end_turnmeans it has finished. A cap is a safety net, not the way a loop should end. Searching the reply for words is the parsing of natural language the guide names as an anti-pattern, and an extra call to ask costs more and can be wrong too.Question 2
Developer productivity. An agent session spent an hour mapping a legacy billing module. You want to try two refactoring approaches, each starting from that analysis, without one attempt's reasoning leaking into the other. What do you do?
Not this time. Forking gives you a second session that starts from the same history, and the original stays as it was, so both attempts share the hour of analysis and nothing else. Asking one session to set an attempt aside doesn't take it out of its context. Mapping the module twice throws the work away. Starting fresh from a summary is the right call when earlier tool results have gone stale; here they still hold, and a summary loses detail. One thing to know: a fork copies the conversation, not the files, so edits made in either fork land in the same folder.
Question 3
Multi-agent research. Every question goes through all four subagents (search, document analysis, synthesis and report), so "What year was the company founded?" takes six minutes and comes back as a long report. What should change?
Not this time. Deciding which subagents a question needs is the coordinator's job. A one-fact question may need one search and a short answer, not the whole pipeline. Running all four at once still runs all four, and synthesis can't start before there is something to combine. A larger model plans the same fixed route. Stored reports only answer what was asked before, and they go out of date.
Question 4
Customer support. When the order database times out,
lookup_orderreturns an empty list, the same as for a customer with no orders. The agent has told several customers they have never ordered anything. What should the tool return on a timeout?Not this time. An empty list is a real answer: the search worked and found nothing. A timeout is a search that failed, and the agent has to know that to retry, wait or tell the customer the truth. MCP reports a tool failure in the result with
isErrorset, and a category and a retry flag let the agent pick the next step. A line in the description can't say which call failed. A bare "Operation failed" hides whether a retry could help. Retrying without end leaves the customer waiting: a short retry inside the tool is fine, but what it can't fix has to come back as an error.Question 5
Code generation with Claude Code. The team uses Jira and wants every developer's Claude Code to reach it, without anyone's API token going into the repository. What do you set up?
Not this time.
.mcp.jsonin the project is the shared, version-controlled place for servers the whole team uses, and Claude Code expands${VAR}references in it, so the file can be committed while each developer keeps the token in their own environment. For a common tool like Jira, a ready-made community server is a better choice than building your own. A token in code is a secret in the repository.~/.claude.jsonis for personal servers, and copies set up by hand drift apart. An ignored.mcp.jsonis no longer shared.Question 6
Code generation with Claude Code. Before a change, Claude has to find every use of a deprecated date helper across a large repository. Last time, the search output filled so much of the context that the later edits lost track of the plan. What helps most?
Not this time. The Explore subagent is made for noisy searching: it's read-only, does its searching apart from your conversation and returns a summary, so the main conversation keeps room for the plan and the edits.
/clearthrows away the findings with the noise. Reading every file first fills the context even faster than the search did. Plan mode holds back edits until you approve a plan; it doesn't keep search output out of the context.Question 7
Claude Code in CI. Your review job runs again on every new commit to a pull request, and each run posts its findings again, so one pull request collects the same comment five times. What should the job do?
Not this time. The review doesn't know what it said last time unless you tell it. Given the earlier findings and asked for only issues that are new or still not fixed, it stops repeating itself and still catches problems pushed later. Matching text word for word fails as soon as a run words the same issue differently. Reviewing only the first run misses everything after it.
--max-turnslimits the turns within one run; it knows nothing about earlier runs.Question 8
Structured data extraction. A check validates each extracted contract and retries with the error when it fails. For some contracts every retry fails on
hourly_rate: the contract says "rates as set out in Schedule B", and Schedule B is a separate file you don't send. What do you do?Not this time. Retrying with the error works when the model got the format or the structure wrong. It can't work when the fact isn't in what the model was given: no number of tries finds a rate that sits in another document. Send that document, or let the field be empty and pass the case to a person. More retries pay for the same failure again. The error in the system prompt changes nothing, because the rate still isn't there. A stricter schema pushes the model to make up a number, which is worse than an empty field.
Question 9
Claude Code in CI. One session writes the code for a pull request and then reviews it. Its reviews nearly always pass, and the bugs turn up later. What gives the review the best chance of catching them?
Not this time. The session that wrote the code still holds the reasoning behind it, so it tends to agree with its own choices. A separate instance with only the diff and the review criteria has no such pull; Anthropic's Claude Code guidance says the same, that a fresh context improves review because Claude isn't biased toward code it just wrote. An instruction to be critical and extended thinking both keep the same context. A second review in the same session repeats the same blind spots, and keeping only issues found twice drops real ones.
Question 10
Multi-agent research. Two credible sources give different figures for the share of small firms using AI: 34% in a 2024 survey and 41% in a 2025 one. The synthesis agent has to put this in the report. What should it do?
Not this time. Two credible figures that differ are part of the finding. Keeping each with its source and date lets the reader see what each one measured, and the dates show this may be a change over a year rather than a contradiction. An average is a number neither source reported. Quietly keeping only the newer figure hides a source the reader may need, and dropping the figure leaves a gap the report never mentions.
Sources
Where each fact comes from
How to study for it, domain by domain
Study the domains in order of weight: the larger the share, the more questions it sets. Each one links to our own free pages that teach it.
27%
Agentic architecture and orchestration
20%
Claude Code configuration and workflows
20%
Prompt engineering and structured output
18%
Tool design and MCP integration
15%
Context management and reliability
Questions
- Who can take the Claude Certified Architect – Foundations exam?
- Only people who work at organisations in the Claude Partner Network. You register on Anthropic's Partner Academy with a company email on a domain listed in your organisation's partner record, and you need to be 18 or older with a government-issued photo ID. If your employer is not a partner, you cannot register, though you can still practise with this sample.
- How many questions are on the Claude Certified Architect – Foundations exam?
- 60 multiple-choice and multiple-response questions in 120 minutes, each saying how many answers to pick. They come from 4 production scenarios chosen at random from a bank of 6, so two candidates may not see the same scenarios.
- What is the passing score for Claude Certified Architect – Foundations?
- A scaled score of 720 on a range of 100 to 1,000. The report also gives the share you answered correctly in each domain, but pass or fail rests on the total scaled score alone.
- How much does the Claude Certified Architect – Foundations exam cost, and can I retake it?
- $125 an attempt, less any discount for your organisation's partner tier. After a fail you wait 14 days, then 30 after a second fail and 90 after a third, with at most 4 attempts in a rolling 12 months, each at the full fee.
- How do you renew the Claude Certified Architect certification?
- Before it runs out at 12 months, you review what has changed and pass a free, non-proctored assessment on the Partner Academy. If the credential lapses, you take the whole exam again and pay the full fee.
- Are these the real Claude Certified Architect exam questions?
- No. They are our own questions, written to the task statements in Anthropic's exam guide and answered from Anthropic's documentation. The real questions are confidential; candidates accept a non-disclosure agreement before the exam.