claudearchitects.org

Resources & Wiki

The community knowledge base for the Claude certification program — the exam domains and objectives for all four tracks, how to prepare, a glossary, an FAQ, and the official documents.

CCAO-FAssociate · Foundation · Beginner

Claude Certified Associate — Foundations

The entry-level certification for everyday Claude users. Covers configuring Projects, prompting fundamentals, managing context and memory, using Claude's interfaces, and responsible-use principles.

Price
$99 USD
Format
~50 items · ~90 min
Practice bank
316 questions
Delivery
Pearson VUE

Domains & objectives

Projects & Configuration
  • Configure a Claude Project with custom instructions that set role, tone, and explicit output rules.
  • Add and organize knowledge sources so guidance persists across every conversation in a Project.
  • Resolve vague or conflicting instructions into explicit, grounded operational boundaries.
  • Decide what belongs in custom instructions vs the knowledge base vs a single message.
Prompting Fundamentals
  • Write prompts that state context, goal, audience, and the desired output format.
  • Use worked examples (few-shot) to steer output structure and quality.
  • Iteratively refine a prompt to fix format, length, or tone problems.
  • Ask Claude to reason step by step for complex analytical tasks.
Context & Memory
  • Keep long conversations productive with summarization and restated constraints.
  • Recognize context-window limits and decide when to restart, summarize, or persist state.
  • Focus Claude on the relevant section of a large document instead of dumping everything.
  • Start a clean conversation seeded with a concise summary when a thread becomes cluttered.
Claude Features & Interfaces
  • Choose the right Claude surface for a task (chat, Projects, artifacts, file upload).
  • Explain what custom instructions, the knowledge base, and a single message each do.
  • Set standing preferences (language, tone) at the Project level rather than per message.
Responsible Use
  • Apply responsible-use principles: avoid deception, respect privacy, avoid harmful content.
  • Decide when a human must review a high-stakes output before it is acted on.
  • Recognize and avoid fabrication/hallucination, and verify facts that matter.
CCDV-FDeveloper · Foundation · Easy

Claude Certified Developer — Foundations

For developers building on Claude. Covers the Messages API, the Claude Agent SDK, tool use and MCP, prompt engineering for code, streaming, and evaluation and testing. A good place to start if you're new to building on the platform.

Price
$125 USD
Format
~60 items · ~120 min
Practice bank
678 questions
Delivery
Pearson VUE

Domains & objectives

Messages API
  • Use the system prompt for application-level, persistent instructions.
  • Maintain multi-turn context by resending conversation history (the API is stateless).
  • Use streaming to keep interactive apps responsive.
  • Apply prompt caching for large, stable prefixes to cut cost and latency.
Agent SDK
  • Build agents with the Claude Agent SDK using built-in tools (Read, Write, Bash).
  • Use a coordinator that owns shared state and passes findings to subagents.
  • Scope an agent's tools and permissions to the minimum its task requires.
Tool Use & MCP
  • Define tools with clear names, precise descriptions, and well-typed input schemas.
  • Run the tool-use loop correctly: return a tool_result referencing the tool_use id.
  • Use MCP to connect models to external tools and data via interoperable servers.
  • Configure tool_choice to guarantee a tool is called when structured output is required.
Prompt Engineering for Code
  • Produce reliable structured output using a JSON schema, examples, and tool use.
  • Ground the model in real docs/types and verify with tests to avoid hallucinated APIs.
  • Write prompts that keep code output scoped, correct, and easy to parse.
Evaluation & Testing
  • Build representative test sets with expected outcomes or checkable criteria.
  • Prefer deterministic, programmatic checks (schema validation, assertions) where possible.
  • Test edge cases, not just the happy path, to catch regressions.
CCAR-FArchitect · Foundation · Hard

Claude Certified Architect — Foundations

The deeper foundation exam — architecture, design decisions and the trade-offs behind them. Covers Claude Code and developer tooling, prompt and context engineering, agentic patterns, integration, and governance foundations.

Price
$125 USD
Format
~60 items · ~120 min
Practice bank
456 questions
Delivery
Pearson VUE

Domains & objectives

Multi-Agent Orchestration & Subagents
  • Construct subagent prompts that include all findings, structured data, and source metadata needed without returning to the coordinator.
  • Evaluate subagent delegation strategies (goal-oriented vs procedural) balancing adaptive behavior with coordinator visibility.
  • Configure subagent invocations with tool restrictions, context scoping, and role-constraining system prompts.
  • Diagnose misconfigured subagent spawning (missing tool permissions, wrong AgentDefinition params, absent coordinator-to-subagent wiring).
  • Evaluate orchestration patterns (coordinator-worker, parallel, sequential pipelines) for coverage, latency, and reliability.
  • Distribute tools so each subagent gets only the tools its role requires.
Agentic Workflows & Task Design
  • Configure agentic loops to emit multiple tool calls in one turn for parallel execution of independent subtasks.
  • Decompose complex tasks into dynamically generated subtasks that adapt to new findings.
  • Select the agentic review architecture (plan mode, direct execution, multi-phase) by scope, risk, and approval needs.
  • Structure iterative refinement with concrete input-output examples, targeted feedback, and batched issues.
Claude Code Configuration & Tooling
  • Configure Claude Code CLI for CI/CD using non-interactive flags, permission modes, and cost/turn limits.
  • Design Claude Code review configurations that load project standards, restrict tools, and emit structured output.
  • Apply the context: fork frontmatter to Skills/slash commands for isolated subagent context.
  • Select the right configuration mechanism (CLAUDE.md, .claude/rules/ globs, Skills, hooks, settings permissions).
  • Select the right built-in tool (Grep, Glob, Read, Bash) for a codebase task.
  • Improve automated test generation with existing tests as context and criteria for meaningful assertions.
Context & Session Management
  • Apply session resumption (targeted re-analysis of changed files, context injection) without repeating work.
  • Design state persistence for multi-agent pipelines that resume after interruption.
  • Apply systematic codebase exploration (Grep/Glob/Read) while managing context-window limits.
  • Apply context management (subagent isolation, scratchpad files, targeted reading) across long sessions.
  • Apply context-window optimization (summarization, sliding windows, structured state, selective retention).
Structured Output & Tool Use
  • Select the most reliable structured-output method (tool use with JSON schema, prompt formatting, prefilled responses).
  • Design extraction schemas with optional/nullable fields and enums to represent missing or ambiguous data without fabricating.
  • Implement tool use with JSON schemas and configure tool_choice to guarantee invocation.
  • Resolve structured-output truncation by splitting scoped calls and merging results, not just raising max_tokens.
  • Sequence multi-tool workflows so prerequisite data is obtained before dependent tools run.
MCP & API Integration
  • Distinguish MCP resources from tools and expose content as resources to reduce exploratory calls.
  • Integrate MCP servers (server scope, environment-variable auth, tool discovery verification).
  • Write MCP tool descriptions that distinguish purpose, inputs, boundaries, and relationships.
  • Select the API processing mode (synchronous Messages API vs asynchronous Message Batches API).
Review, Extraction & Evaluation
  • Design human-review routing by confidence, document characteristics, and field ambiguity rather than random sampling.
  • Design specialized review passes that separate concerns (security, business logic, API design) with dedicated few-shot examples.
  • Apply extraction accuracy patterns (structured schemas, format normalization, few-shot) to reduce hallucination.
  • Reduce automated-review false positives with project conventions and exclusion criteria as persistent context.
  • Design synthesis behavior that preserves source-level uncertainty instead of collapsing conflicting data.
CCAR-PArchitect · Professional · Advanced

Claude Certified Architect — Professional

The advanced architect certification. Covers designing complex multi-agent systems, production scale and reliability, evaluation strategy, cost and performance trade-offs, and enterprise governance.

Price
$175 USD
Format
~63 items · ~120 min
Practice bank
309 questions
Delivery
Pearson VUE

Domains & objectives

Solution Design & Architecture
  • Translate a business problem into a Claude-based AI solution.
  • Design an end-to-end architecture (input, processing, output, feedback loops).
  • Select an architectural pattern (workflow, agentic, augmented LLM) for a scenario.
  • Design multi-agent systems and orchestration strategies.
  • Align a solution to business value pillars (efficiency, transformation, productivity, cost, SLAs).
Models, Prompting & Configuration
  • Select an appropriate Claude model based on capability, cost, and latency trade-offs.
  • Apply prompt engineering techniques (zero-shot, few-shot, chain-of-thought).
  • Evaluate tool/agent configuration for capability bloat.
  • Justify accuracy-latency trade-offs in a configuration decision.
Security, Compliance & Governance
  • Analyze authentication and authorization requirements to identify security gaps.
  • Evaluate connection protocols and select the appropriate integration mechanism.
  • Ensure compliance with regulations such as GDPR, HIPAA, and FedRAMP.
Safety & Risk
  • Implement guardrails and safety controls.
  • Identify risks, limitations, and failure modes of LLM systems.
  • Apply human-in-the-loop validation strategies.
  • Address ethical AI considerations (bias, fairness, transparency).
Evaluation & Testing
  • Design evaluation datasets and test frameworks using mixed methodologies.
  • Conduct A/B testing and iterative improvements.
  • Diagnose system issues (prompt failure, hallucinations, model mismatch).
Observability & Optimization
  • Analyze observability challenges and select monitoring strategies at scale.
  • Optimize token usage, latency, and cost-performance trade-offs.
  • Monitor system performance using logging and observability tools.
Delivery & Stakeholder Management
  • Conduct structured discovery and requirement gathering.
  • Communicate architectural decisions and trade-offs.
  • Manage stakeholder feedback loops and expectations, including SLAs.
  • Document architectures and provide implementation guidance.
  • Support lifecycle phases (discovery, design, handoff, monitoring, iteration).
Developer Enablement
  • Configure Claude tools and environments for teams (e.g., Claude Code).
  • Improve developer workflows using AI-assisted tooling.
  • Support debugging and operational issue resolution.

How to prepare

  1. Take the free Anthropic Academy courses for your track (see below) to cover the concepts end to end.
  2. Study the domains & objectives above — they map directly to what each exam tests.
  3. Drill our free mock exams in Practice mode (instant explanations), then simulate the real thing with a timed Mock exam and review the per-domain score report.
  4. Read the official documents (infographic, policy, terms) before you book.

FAQ

Are these official exams?

The certifications are Anthropic's official role-based credentials, delivered proctored via Pearson VUE with a Credly badge. This site is an independent, community-built study resource — not affiliated with Anthropic.

How much do they cost?

Approximately: Associate $99, Developer $125, Architect Foundations $125, Architect Professional $175 (USD, paid to Anthropic). Always confirm current pricing with the vendor.

What's the passing score?

The reports use a 720 / 1000 scale (about 72%). The pass mark used in our mock exams is a site default, not a published vendor cut score.

How do I register?

Through the Claude SG partner network: complete the screening, receive a free @claudecode.sg partner-network email, sign a short agreement, then sit the proctored exam. See the Certification page.

How long is a certification valid?

Around 12 months (verify with the vendor).

How should I prepare?

Take the recommended free Anthropic Academy courses, study the objectives on this page, then drill our free mock exams with explanations.

Glossary

Claude Projects
A workspace with custom instructions and a knowledge base that persist across every conversation in the Project.
Custom instructions
Standing guidance (role, tone, rules, output format) applied to all chats in a Project.
Knowledge base
Reference documents attached to a Project so Claude can use them across conversations.
Context window
The amount of text (tokens) Claude can consider at once; managing it well is core to long tasks.
System prompt
Application-level, persistent instructions sent with every API request.
Messages API
The stateless API for multi-turn conversations; you resend history each request.
Message Batches API
Asynchronous processing for large, non-latency-sensitive workloads.
Tool use
Letting Claude call defined tools (with name, description and a typed input schema) and act on the results.
MCP (Model Context Protocol)
An open standard connecting models to external tools and data via interoperable servers.
Claude Agent SDK
The SDK for building agents with built-in tools (Read/Write/Bash), subagents and a coordinator.
Subagents
Specialized agents a coordinator dispatches; the coordinator owns shared state across them.
Prompt caching
Reusing a large, stable prompt prefix across requests to cut cost and latency.
Structured output
Enforcing a schema (via tool use / JSON) so responses can be parsed reliably.
Claude Code
Anthropic's agentic coding tool, configurable via CLAUDE.md, rules, hooks, skills and settings.

Downloads & official links