Claude certification glossary: one plain definition for every term you meet on the way to a Claude Certified Architect, Developer or Associate credential - the exam codes and their spellings, how the exam is scored and booked, what the domains test, and the words this prep site uses. Each entry links to the page that goes deeper.
Anthropic's architect-track credential family: Foundations (CCAR-F) and Professional (CCAR-P). There is no separate Claude solution architect exam.
AlsoClaude architect certificationClaude solution architect certificationClaude AI architect
The four proctored Anthropic exams—CCAO-F, CCAR-F, CCDV-F and CCAR-P—plus free course-completion certificates available through Anthropic Academy.
AlsoClaude certificationsClaude certificateAnthropic Claude certificationClaude Anthropic certification
The Foundations architect exam: 60 questions, 120 minutes, 5 domains, 720 of 1000 to pass, $125 list price and 12-month validity.
AlsoCCA-FCCAFCCARFClaude Certified Architect - Foundations
The Professional architect exam: 63 questions, 120 minutes, 7 domains, 720 to pass, $175 list price and 12-month validity.
AlsoCCA-PCCARPClaude Certified Architect - Professional
The developer-track exam: 53 questions, 120 minutes, 8 domains, 720 to pass, $125 list price and 12-month validity.
AlsoCCD-FCCDVFCCTV-FClaude Certified Developer - Foundations
The non-technical associate exam: 60 questions, 120 minutes, 7 domains, 720 to pass, $99 list price and 12-month validity.
AlsoCCA-ACCAOCCAOFClaude Certified Associate - Foundations
Anthropic's certification programme and learning portal. The credentials are earned through proctored exams, while the supporting courses and course certificates are free.
AlsoAnthropic certificationsAnthropic certifiedAnthropic Academy
Anthropic Academy hosts courses and exam-access requests, while Anthropic's learning site lists the certifications. This website is independent preparation, not the vendor.
AlsoClaude Certified Architect official websiteofficial Claude certification page
The learning platform Anthropic uses for Academy courses and certification access requests. Current eligibility details are dated on the registration page.
AlsoAnthropic Academy portalanthropic.skilljar.com
Anthropic's partner programme. Eligible cohorts may receive a fully covered exam attempt; the registration page carries the dated eligibility rule.
Alsopartner networkpartner cohort
A verifiable digital badge issued after a successful certification result, with a public credential record that can be claimed and shared.
Alsodigital badge
A course certificate proves course completion; a certification proves a proctored exam pass. Anthropic offers both, but they are different credentials.
Alsocourse certificateClaude Code Essentials certificateClaude 101
Academy courses and course certificates are free, and eligible partner cohorts may receive an exam attempt at no cost. Public exam list prices remain separate.
AlsoClaude free certificationClaude certification freefree Claude certifications
Anthropic's blueprint document for each certification, covering domains, weights, format and sample scenarios. This site links the available PDFs.
Alsoexam guide PDFClaude exam guidesyllabusstudy guide
A scored area of an exam with a defined weight, such as CCAR-F domain 1, Agentic Architecture and Orchestration, at 27 percent.
Alsodomainblueprint domain
The share of scored questions assigned to a domain. CCAR-F uses weights of 27, 18, 20, 20 and 15 percent.
Alsoweighting
A result reported on a 1000-point scale. A score of 720 passes; it should not be read as a raw percentage.
Also1000-point scale720
The minimum scaled result needed to pass: 720 of 1000 for every current Claude certification exam.
Alsopassing scorecut score720
A question built around a situation, constraints and four options with one best answer. It tests judgment rather than recall alone.
Alsoscenario-based questionCCA-F scenario
The dated process of requesting access through Skilljar, completing eligibility checks and then scheduling the proctored exam.
Alsoregisterattestation formaccess requesthow to book
Current list prices are $125 CCAR-F, $175 CCAR-P, $125 CCDV-F and $99 CCAO-F, in US dollars.
Alsoexam feelist priceClaude certification costClaude certification price
The delivery vendor named in the exam guides. Candidates sit through online proctoring or at an available test centre under programme rules.
AlsoOnVUEtest centre
An exam taken remotely under camera, microphone and screen monitoring. Identity, room, desk and device requirements apply before the session starts.
AlsoOnVUEremote proctoring
Another exam attempt after a failed result. The registration page carries the dated waiting periods, annual attempt limit and applicable fee.
Alsoretake policyattempts
How long a passed credential remains current before renewal. Every certification in the present registry has a 12-month validity period.
Alsorecertification12 months
The identity, environment, device and conduct requirements that determine whether a proctored exam can start and remain valid.
Alsoexam day
Tempting answer patterns that fail under the exam's stated constraints. The anti-pattern guide groups them by domain and explains the correction.
Alsoexam traps
A preparation schedule that assigns reading, practice and mocks according to domain weight and available time.
Also4-week planhow to prepare
Copies of real exam items obtained outside authorized materials. Blueprint-aligned practice questions are the legitimate way to rehearse the format.
Alsoleaked questions
A free ten-question check across the selected certification that reports a score, identifies the weakest domain and recommends the next study step.
Also10-question diagnostic
Original questions written to the exam blueprint, used to practise the format, domain coverage and reasoning without reproducing real exam items.
Alsopractice testpractice questionsClaude practice exammock exam questions
A timed, scored simulation using the official domain weights. Three attempts are included, with additional attempts unlockable using earned XP.
Alsopractice mockfinal mock
A short public set of certification questions with immediate answers and explanations, designed to show how the exam reads before a longer attempt.
Alsosample questions
A focused set of questions from the weakest diagnosed domain, used before a mock to rehearse the area most likely to suppress the score.
Alsoweak-domain warm-up
A study sequence shaped by diagnostic and mock evidence, prioritising weaker and higher-weighted domains instead of assigning the same path to everyone.
Alsopersonalised plan
Different names for structured preparation. The free course hub combines Anthropic lesson mirrors, focused deep dives and links into concepts and practice.
AlsoClaude certification courseClaude Certified Architect trainingbootcamp
A learning-progress signal earned through focused reading and correct practice. It indicates retained work on this site, not a guaranteed exam result.
Alsodomain mastery
A composite of learning activity used for progress and league placement. The exam-relevant result remains the mock score against the 720 pass mark.
Alsoactivity score
Consecutive days with qualifying study activity. A streak measures consistency and contributes to platform progress, but it is not an exam-readiness forecast.
A weekly group of learners ranked by sparks earned since Monday, with promotion and drop places at the reset.
Alsoweekly tier
Short recall prompts reviewed over time. The guide explains how cards complement concepts, questions and mock analysis.
Alsospaced repetition
A focused explanation of one exam primitive, with examples, traps, supporting sources and practice that connect the term to exam decisions.
Alsoconcept page
A production architecture that combines several concepts into one constrained problem, then traces the design, implementation and common failure modes.
Alsoscenario deep dive
A concise study version of an Anthropic Academy course, retaining the lesson sequence while linking each takeaway to relevant exam material.
AlsoSkilljar mirror
A compact lookup for rules, distractors and comparisons when a learner needs the answer quickly rather than another full lesson.
Alsoquick reference
A small source-quality marker attached to a claim, helping learners distinguish official material from secondary, community or disputed evidence.
Alsosourcing marker
A controlled multi-turn cycle that reads a model response, dispatches tools, appends observations and branches on the structured stop reason.
Alsoagent loop
A specialised worker given an isolated context and bounded task by a coordinating agent, returning a compact result rather than its full history.
Alsosub-agent
The explicit transfer of selected facts or results between isolated agents without merging their entire contexts.
Alsostate handoff
The structured response field that tells an agent loop whether Claude finished, requested a tool or stopped because of an output limit.
Alsoend_turntool_usemax_tokens
The contract that lets Claude request a named external capability with structured input while application code performs the actual operation.
Alsofunction calling
The API control that allows, requires or constrains tool use without removing the application's duty to validate arguments and results.
Alsotool_choice
Anthropic's open protocol for exposing tools, resources and prompts to AI clients through a standard discoverable interface.
AlsoMCP serverMCP clientMCP toolMCP resource
Deterministic lifecycle handlers that inspect, block or react to tool activity outside the model's probabilistic reasoning.
AlsoPreToolUsePostToolUse
The precedence and scope of persistent Claude Code instruction files across enterprise, user and project locations.
Alsoproject CLAUDE.mduser CLAUDE.mdenterprise CLAUDE.md
Repository or environment files that give coding agents durable context, commands, constraints and conventions before they work.
AlsoCLAUDE.mdAGENTS.md
A read-only working mode used to inspect a codebase, clarify the approach and sequence changes before editing files.
Alsoplanning mode
Reusable task-specific instructions and resources loaded when relevant, keeping specialised procedures separate from permanent repository context.
Alsoagent skills
Practices for preserving verified progress, resuming work and limiting context drift across long or interrupted coding sessions.
Alsosession checkpointresume session
The highest-priority application instruction that defines Claude's role, boundaries, output expectations and persistent operating context.
Alsosystem instruction
Reusing a stable prompt prefix across API calls to reduce repeated processing cost and latency when the cached content remains identical.
Alsocache breakpoint
An asynchronous API for large groups of independent requests when lower cost matters more than immediate responses.
Alsomessage batches
Model output constrained to a defined schema so application code can validate and consume it without relying on brittle text parsing.
AlsoJSON schema output
Delivering response events incrementally so an interface can show progress before the complete model output is available.
Alsoserver-sent eventsSSE
Methods for clarifying instructions and context through structure, examples and reasoning controls selected for the task.
AlsoXML tagsfew-shot promptingextended thinking
The bounded input history available to the model and the process of summarising older material to preserve useful state within that limit.
Alsocontext limitcompaction
Anthropic's AI-fluency model for deciding what to delegate, describing the task, evaluating the output and applying responsible judgment.
Alsodelegationdescriptiondiscernmentdiligence
A compact, re-read set of authoritative facts kept separate from the growing conversation so an agent does not drift from the case.
Alsoimmutable case facts
Untrusted content that tries to alter model instructions or cause unauthorized actions, requiring isolation, least privilege and deterministic controls.
Alsoindirect prompt injection
Layered controls that constrain inputs, tools, outputs and escalation paths instead of relying on the model to police itself.
Alsosafety controls
Logs, traces, metrics and evaluations that reveal what an AI system did, why it failed and whether behavior changes over time.
Alsoproduction monitoring
Choosing a model by capability, latency, cost and risk requirements rather than defaulting to the largest option for every task.
AlsoOpusSonnetHaiku
A plausible model statement unsupported by the available evidence, requiring verification rather than confidence-based acceptance.
Alsoconfabulation
Following wording mechanically while missing the intended outcome or safety constraint, a common distractor pattern in scenario questions.
Alsoliteral compliance
Matching reliance on an AI result to the evidence, stakes and available verification rather than trusting or rejecting it categorically.
Alsoappropriate reliance
The context currently available to a model during a task, distinct from durable application storage or a person's biological memory.
AlsoAI working memory
A credential intended to demonstrate skill in designing or building AI agents. The relevant choice depends on the platform and role.
Alsoagent certification
There is no standalone Model Context Protocol exam; MCP skill is assessed inside Claude architect and developer certifications.
AlsoModel Context Protocol certification
There is no separate Claude Code exam. Claude Code configuration and workflows are tested inside the Claude certification portfolio.
AlsoClaude Code certificate
A narrower credential focused on prompting, unlike architecture certifications that also test tools, agents, operations, safety and reliability.
Alsoprompt engineer certificate
A broad engineering credential covering model and application work. Claude certifications are more tightly aligned to Anthropic's products and role tracks.
AlsoAI engineer certificationLLM certification
A credential for end-to-end AI system design. Vendor-specific versions differ in platform primitives, delivery model and assessed architecture decisions.
Alsoartificial intelligence architect certification
Cloud-platform credentials that assess services and architectures in their own ecosystems, compared with Anthropic's Claude-specific certification portfolio.
Alsocloud AI certification
Free learning material for using Claude and its tools. Course completion supports preparation but does not equal a proctored certification pass.
AlsoClaude courseClaude 101
CCAF commonly means Community College of the Air Force, which is unrelated. On this site, CCA-F is shorthand for the CCAR-F exam.
AlsoCommunity College of the Air Force
CCA can mean Cloudera Certified Associate or a car-battery rating. Here, CCA-F refers to Claude Certified Architect - Foundations.
AlsoCloudera Certified Associatecold cranking amps
CCDV has unrelated meanings and CCTV-F is a common mistyping. The Claude developer exam code is CCDV-F.
AlsoCCDVCCTV-F
MCP can mean Microsoft Certified Professional. In Claude architecture material, MCP means Model Context Protocol.
AlsoMicrosoft Certified Professional
