Claude Certified Architect - Foundations Study Guide
The Certifyforce Claude Certified Architect - Foundations study guide covers the official Claude Certified Architect - Foundations exam outline in 30 units across all 5 exam sections, one unit per exam-guide topic. The exam itself has 60 scored questions, a 120-minute limit and a passing score of 720 (scaled 100–1,000), and its largest section is Agentic Architecture & Orchestration at 27%. Every unit is verified against official Anthropic documentation, cites its sources, and links to the practice questions that test it. Last updated October 22, 2026.
Focused on designing and building production Claude solutions with the Claude Agent SDK, Claude Code, the Claude API and MCP: agentic architecture and orchestration, tool design and MCP integration, Claude Code configuration, prompt engineering and structured output, and context management and reliability.
30 lessons · 5 exam sections · about 4 h of reading ·
Exam sections
Unlock the full guide
Every lesson's summary, exam expectations, key facts and self-check questions are free to read. The Claude Certified Architect - Foundations pack unlocks the self-check answers, core concepts, exam traps and worked scenarios, reading progress, personalized “read next”, and unlimited practice exams.
Agentic Architecture & Orchestration
27% of the exam- Agentic loops and stop_reason control~8 minHow a client-tool agentic loop works on the Messages API and in the Agent SDK: continue on stop_reason tool_use, branch on every other value, append full assistant turns and all tool_result blocks, and keep turn and budget caps as backstops that report the run as incomplete.
- Coordinator-subagent orchestration~8 minHow a coordinator agent routes, scopes and checks the work of specialist subagents: hub-and-spoke communication, description-based selection, effort scaling, scope partitioning, the gap-filling refinement loop, when multi-agent is worth its token cost, and how to cap and trace the agent tree.
- Spawning subagents and passing context~8 minWhat crosses the boundary when a coordinator spawns a subagent with the Task tool (now named Agent): only the delegation prompt, so findings, constraints and quality criteria must be written into it. Covers AgentDefinition fields, forks as the exception, structured hand-offs that keep source metadata, and parallel spawning in one response.
- Workflow enforcement gates and handoffs~8 minHow to enforce workflow order in code instead of prompts: prerequisite gates with PreToolUse and PostToolUse hooks, handler checks or a withheld tool, the Agent SDK permission evaluation order, human approval pauses with defer, multi-concern decomposition, and structured handoff summaries for human agents.
- Agent SDK hooks for interception and normalization~8 minHow Agent SDK hooks turn business rules into code: a PreToolUse hook that denies a policy-violating call and redirects Claude to escalation, a PostToolUse hook that normalizes or redacts tool results before Claude reads them, and how to decide which rules need a hook instead of a prompt.
- Task decomposition: chains vs dynamic plans~8 minHow to split complex work for Claude: fixed prompt chains for predictable multi-aspect reviews, parallel sections for independent checks, per-file passes plus a cross-file integration pass for large reviews, and adaptive plans that map, prioritize and re-plan for open-ended tasks.
- Sessions: resume, fork or start fresh~8 minWhen to resume a named or saved session, when to fork it into independent branches, and when to start fresh with a structured summary instead, with the CLI and Agent SDK spellings side by side and what a resumed or forked session does and does not restore.
Tool Design & MCP Integration
18% of the exam- Tool descriptions, boundaries and naming~8 minHow Claude picks a tool from its name and description, and how to fix misrouting: descriptions that state what, when, inputs and limits; renaming and namespacing overlapping tools; splitting generic tools or consolidating endpoint wrappers; and removing system-prompt wording that overrides good descriptions.
- Structured error responses for MCP tools~8 minHow tools report failure so an agent can recover: MCP isError results versus JSON-RPC protocol errors, is_error in the Messages API and the Agent SDK, error categories with isRetryable and a readable message, business-rule refusals, expired handles, and why a failed query must never come back as an empty result.
- Distributing tools and configuring tool_choice~8 minWhy each agent should hold only its role's tools, how to scope them with tools, disallowedTools and inline mcpServers (and why allowedTools only auto-approves), and every tool_choice value: auto, any, a forced tool and none, including forcing a first step and the models that reject forced choice.
- MCP servers in Claude Code and agents~8 minWhere to configure an MCP server (local, project .mcp.json or user scope), how ${VAR} expansion keeps tokens out of git, why every configured server's tools are usable together, when to expose data as MCP resources instead of tools, and how Agent SDK agents get permission to call MCP tools.
- Built-in tools: Read, Write, Edit, Bash, Grep, Glob~8 minGrep for file contents versus Glob for file paths, Read with offset and limit for large files, Edit's unique-match rule and the Read + Write fallback, incremental exploration from Grep entry points, tracing wrapper modules by exported name, and why Glob and Grep may be absent by default.
Claude Code Configuration & Workflows
20% of the exam- CLAUDE.md hierarchy, imports and scoping~8 minWhere each CLAUDE.md lives (managed, user, project, local, subdirectory), how the files concatenate instead of overriding, what @path imports and the .claude/rules/ directory do and do not save, and how to diagnose loading with /memory, /context and claudeMdExcludes.
- Custom slash commands and skills~8 minHow .claude/commands/ files and SKILL.md skills create the same /name commands, where project and personal skills live and which wins a name clash, and what argument-hint, context: fork, allowed-tools, disallowed-tools and disable-model-invocation actually do. Also when a workflow belongs in a skill rather than CLAUDE.md.
- Path-specific rules in .claude/rules~8 minHow a paths glob in a .claude/rules/ file makes conventions load only when Claude reads, writes or edits a matching file, how the patterns resolve, what silently breaks scoping, and when a path-scoped rule beats a subdirectory CLAUDE.md. Also covers compaction, symlinked rules and logging loads with the InstructionsLoaded hook.
- Plan mode vs direct execution~8 minWhen to plan first and when to just make the change in Claude Code: the one-sentence-diff test, entering and defaulting to plan mode, the approval options and Ctrl+G, plan-then-execute, and the built-in Explore and Plan subagents that keep verbose discovery out of the main context.
- Iterative refinement: examples, tests and interviews~8 minHow to steer Claude Code toward the right result over several rounds: concrete input/output examples when prose is misread, test-first iteration without editing or hard-coding to the tests, the interview-then-spec-then-fresh-session pattern, batching interacting issues versus fixing independent ones in turn, and resetting with /clear after repeated failed corrections.
- Claude Code in CI/CD pipelines~8 minRunning Claude Code unattended in CI: -p print mode, --output-format json with --json-schema and reading structured_output, failing on exit codes and missing output, CLAUDE.md as CI project context versus --bare, prior findings and existing tests in context, and the GitHub Actions and GitLab CI/CD integrations.
Prompt Engineering & Structured Output
20% of the exam- Explicit criteria to cut false positives~8 minWhy vague review and classification instructions such as "be conservative" or "only report high-confidence findings" fail, and how categorical report/skip lists, severity definitions with code examples, evidence rules, REVIEW.md and a labeled evaluation set raise precision without hiding real issues.
- Few-shot examples for consistent output~8 minWhen and how to add few-shot examples: format demonstrations, targeted ambiguous cases with reasoning, acceptable-versus-genuine pairs, varied document structures and informal values left null, diverse examples in <example> tags, worked examples, and tool input_examples.
- Structured output with tool use and JSON schemas~7 minHow strict tool use and JSON outputs (output_config.format) guarantee schema-valid structure but not correct values, how to read tool_use input, and how to design extraction schemas with required, nullable, "other" and "unclear" fields plus prompt normalization rules.
- Validation-retry loops for extraction quality~8 minHow to catch the semantic errors strict schemas let through, feed specific validation errors back to Claude, recognise when a retry cannot help because the data is absent, and design schema fields and finding metadata that surface discrepancies and false-positive patterns.
- Message Batches API and batch strategy~8 minWhen the Message Batches API's 50% discount is worth its up-to-24-hour window, how to correlate results by custom_id, resubmit only failed requests, size submission intervals against an SLA, and why client tools cannot run inside a batched request.
- Multi-instance and multi-pass review~8 minWhy a session reviewing its own work misses errors, how to give an independent reviewer only the artifact and the criteria, and how verification passes, rubric-based evaluators and multi-instance voting turn review findings into reliable, routable signals.
Context Management & Reliability
15% of the exam- Preserving critical facts in long conversations~8 minHow to keep amounts, dates, order IDs and rules intact as a conversation grows: pin facts in a case-facts block outside summarized history, trim tool output before it piles up, order long inputs for position effects, and choose between context editing, compaction and notes files.
- Escalation and ambiguity resolution~8 minWhen a support or developer agent should hand off to a human, offer to resolve, or ask a clarifying question: the three escalation triggers, why sentiment and self-reported confidence fail as signals, explicit criteria with few-shot examples, and AskUserQuestion for ambiguity in Agent SDK apps.
- Error propagation across multi-agent systems~8 minHow subagent failures should reach the coordinator: recover transient errors locally, propagate the rest with failure type, attempted query, partial results and alternatives, tell access failures from valid empty results, resume partial subagents, and annotate coverage gaps instead of suppressing errors or aborting the run.
- Context management in large codebase exploration~8 minHow to keep long codebase explorations precise: spot context degradation, persist findings in scratchpad and state files, push verbose searches into subagents, seed each phase with a summary, recover crashed runs from manifests, compact with a focus and block generated code from reads.
- Human review workflows and confidence calibration~8 minWhen it is safe to cut human review of extractions: segment-level accuracy instead of one aggregate number, field-level confidence calibrated on labeled data, routing scarce reviewers to low-confidence and contradictory records, stratified sampling of auto-approved output, and calibrating LLM graders against experts.
- Information provenance in multi-source synthesis~8 minHow multi-agent research keeps every claim traceable: structured claim-source mappings with dates that synthesis merges, conflicting figures annotated rather than resolved by guesswork, established versus contested findings, format per content type, and the API's Citations and search_result blocks.
Common questions about the Claude Certified Architect - Foundations study guide
Is the Claude Certified Architect - Foundations study guide free?
Partly. Every unit's summary and its "What the exam expects", "Key facts" and "Check yourself" sections are free to read without an account. The "Core concepts", "Exam traps" and "Worked scenario" sections, reading progress and unlimited timed mock exams come with the Claude Certified Architect - Foundations pack (USD $19).
How current is the Claude Certified Architect - Foundations study guide?
It is written to the current official Claude Certified Architect - Foundations exam outline, and each unit is checked against Anthropic documentation before it is published. The most recent update was October 22, 2026, and every unit shows its own last-updated date.
How does the study guide map to the official Claude Certified Architect - Foundations exam guide?
There is one unit per topic in the official exam guide, grouped under its 5 weighted sections: Agentic Architecture & Orchestration (27%), Tool Design & MCP Integration (18%), Claude Code Configuration & Workflows (20%), Prompt Engineering & Structured Output (20%), Context Management & Reliability (15%). Each unit quotes the exam objective it covers word for word.
How do the practice questions link to the study guide?
Each unit lists the reviewed practice questions that test it (300 questions are linked across the guide), so you can read a topic and practice it in one click. After a practice exam, every missed question links back to the unit that teaches it.
Certifyforce is an independent study resource and is not affiliated with, endorsed by or sponsored by Anthropic. Claude is a trademark of Anthropic.