Claude Certified Architect - Foundations
Make sound production tradeoffs across Claude Code, the Claude Agent SDK, the Claude API, and MCP.
Developers and architects who design production Claude systems and can reason through ambiguous, scenario-based architecture decisions.
60
Questions
120 min
Time limit
$125
Exam fee
720 / 1000
Passing score
Scenario-based multiple-choice and multiple-response; four scenarios drawn from six
Format
Proctored online or at a test center
Delivery
12 months
Validity
v1.0 · July 2026
Guide
Exam domains
Weighted objectives straight from the exam guide.
Agentic Architecture and Orchestration
27%- Implement the tool-use lifecycle from stop reason to tool result
- Choose coordinator and subagent topologies
- Use Task, allowedTools, isolated context, parallelism, and forked sessions
- Enforce prerequisites, decompose concerns, and structure handoffs
- Use hooks for deterministic controls and normalization
- Choose sequential or adaptive decomposition
- Resume or fork without carrying stale context
Tool Design and MCP Integration
18%- Write precise tool descriptions and boundaries
- Return structured and retry-aware MCP errors
- Manage tool distribution and tool choice
- Scope MCP configuration and resources
- Choose Claude Code file, shell, and search tools safely
Claude Code Configuration and Workflows
20%- Design CLAUDE.md hierarchy, imports, rules, and memory
- Create project and user commands and Skills
- Apply path-specific rules
- Choose plan mode, direct execution, or Explore
- Use examples, test-driven iteration, and interviews
- Run headless CI with structured output and independent review
Prompt Engineering and Structured Output
20%- State explicit evaluation criteria
- Use few-shot examples for consistent judgment
- Enforce schemas with tool use and tool choice
- Validate, retry, and feed back semantic errors
- Use batch processing for appropriate workloads
- Separate generator and independent reviewer passes
Context Management and Reliability
15%- Place facts to reduce lost-in-the-middle failures
- Escalate ambiguity and policy gaps
- Propagate structured errors and partial results
- Use scratchpads, subagents, manifests, and compaction for large codebases
- Calibrate confidence and human review
- Track provenance, conflicts, dates, and content types
Study path · 21 lessons
Go deeper in the main curriculum
Orchestration Patterns
Compare coordinator, delegation, and parallel patterns
Durable Execution
Handle resume, failure, and long-running state
MCP Tool Poisoning
Treat tool descriptions and results as an attack surface
Reviewer Agent
Build independent evidence-based review
Multi-Session Handoff
Preserve structured state without stale context