Anthropic Claude Certified Architect

A free, self-paced 29 module training curriculum for Anthropic Claude Certified Architect, built for engineers architecting agentic systems and tooling with Anthropic Claude. Completing every module issues a printable 29.0 contact hour certificate from TECHLEAD 187 LLC.

• 29 modules
• 29.0 contact hours
• 5 domains
• Self-paced
• Free to use

Domains covered

• Agentic Architecture & Orchestration: 7 modules
• Tool Design & MCP Integration: 5 modules
• Claude Code Configuration & Workflows: 6 modules
• Prompt Engineering & Structured Output: 6 modules
• Context Management & Reliability: 5 modules

Module syllabus

1. 1.1 Design and Implement Agentic Loops for Autonomous Task Execution
   Build agent loops that inspect stop_reason, execute tools, and return results to conversation history for model reasoning on next action.
2. 1.2 Orchestrate Multi-Agent Systems with Coordinator-Subagent Patterns
   Design hub-and-spoke multi-agent systems where a coordinator manages inter-agent communication, task decomposition, and result aggregation.
3. 1.3 Configure Subagent Invocation, Context Passing, and Spawning
   Explicitly pass context to subagents via prompt, use the Task tool with allowedTools configuration, and spawn parallel subagents in a single turn.
4. 1.4 Implement Multi-Step Workflows with Enforcement and Handoff Patterns
   Use programmatic hooks for prerequisite gates and structured handoff summaries for human escalation including customer details and root cause analysis.
5. 1.5 Apply Agent SDK Hooks for Tool Call Interception and Data Normalization
   Use PostToolUse hooks to normalize heterogeneous data formats and implement PreToolCall hooks to enforce policy compliance.
6. 1.6 Design Task Decomposition Strategies for Complex Workflows
   Choose between fixed sequential pipelines (prompt chaining) and dynamic adaptive decomposition based on intermediate findings.
7. 1.7 Manage Session State, Resumption, and Forking
   Use --resume for named session continuation, fork_session for divergent exploration branches, and structured summaries for reliable resumption.
8. 2.1 Design Effective Tool Interfaces with Clear Descriptions and Boundaries
   Write tool descriptions that clearly differentiate purpose, expected inputs/outputs, edge cases, and when to use versus similar alternatives.
9. 2.2 Implement Structured Error Responses for MCP Tools
   Use MCP isError flag with structured metadata including errorCategory (transient/validation/permission) and isRetryable boolean.
10. 2.3 Distribute Tools Appropriately Across Agents and Configure Tool Choice
    Restrict each agent to 4-5 role-specific tools, use tool_choice configuration, and avoid cross-specialization misuse.
11. 2.4 Integrate MCP Servers into Claude Code and Agent Workflows
    Configure project-level .mcp.json with environment variable expansion for credentials; leverage MCP resources to reduce exploratory tool calls.
12. 2.5 Select and Apply Built-in Tools Effectively
    Use Grep for content search, Glob for path matching, Read/Write for full file operations, Edit for targeted modifications.
13. 3.1 Configure CLAUDE.md Files with Hierarchy, Scoping, and Modular Organization
    Use user-level, project-level, and directory-level CLAUDE.md hierarchy; apply @import syntax and .claude/rules/ for modular organization.
14. 3.2 Create and Configure Custom Slash Commands and Skills
    Organize commands in .claude/commands/ (shared) and ~/.claude/commands/ (personal); configure skills with SKILL.md frontmatter.
15. 3.3 Apply Path-Specific Rules for Conditional Convention Loading
    Create .claude/rules/ files with YAML frontmatter path fields using glob patterns to load rules only for matching files.
16. 3.4 Determine When to Use Plan Mode vs Direct Execution
    Use plan mode for complex multi-file changes and architectural decisions; direct execution for well-scoped, clear changes.
17. 3.5 Apply Iterative Refinement Techniques for Progressive Improvement
    Provide concrete input/output examples; use test-driven iteration; employ the interview pattern for surfacing design considerations.
18. 3.6 Integrate Claude Code into CI/CD Pipelines
    Use -p (print) flag for non-interactive mode; configure --output-format json for structured findings; include CLAUDE.md context.
19. 4.1 Design Prompts with Explicit Criteria to Improve Precision
    Write specific review criteria defining which issues to report; avoid vague instructions.
20. 4.2 Apply Few-Shot Prompting to Improve Output Consistency
    Provide 2-4 targeted few-shot examples for ambiguous scenarios, demonstrating output format and edge cases.
21. 4.3 Enforce Structured Output Using Tool Use and JSON Schemas
    Use tool_use with JSON schemas for guaranteed schema-compliant output; configure tool_choice appropriately.
22. 4.4 Implement Validation, Retry, and Feedback Loops
    Append specific validation errors on retry; track detected_pattern fields; design self-correction validation flows.
23. 4.5 Design Efficient Batch Processing Strategies
    Use Message Batches API for 50% cost savings and up to 24-hour processing; match API to workflow latency.
24. 4.6 Design Multi-Instance and Multi-Pass Review Architectures
    Use independent review instances; split large reviews into per-file local analysis plus cross-file integration passes.
25. 5.1 Manage Conversation Context to Preserve Critical Information
    Extract transactional facts into persistent "case facts" blocks; trim verbose outputs; place key findings at beginning.
26. 5.2 Design Effective Escalation and Ambiguity Resolution Patterns
    Escalate when customers explicitly request it or when policy is ambiguous; ask for additional identifiers on multiple matches.
27. 5.3 Implement Error Propagation Strategies Across Multi-Agent Systems
    Return structured error context including failure type, attempted query, and partial results; distinguish access failures from valid empty results.
28. 5.4 Manage Context Effectively in Large Codebase Exploration
    Use scratchpad files to persist findings; spawn subagents for isolated output; maintain structured state for crash recovery.
29. 5.5 Design Human Review Workflows and Confidence Calibration
    Implement confidence scores alongside findings; route reviews based on calibrated confidence; enable developer overrides.

What each module includes

An executive lesson briefing with five key concepts and official vendor documentation references, a scenario based knowledge check scored for mastery, an applied scenario evaluation, and supplemental verified videos plus official free hands on labs where available.

Other curricula

Published by TECHLEAD 187 LLC. Training certificates evidence completed instruction and contact hours; they are not vendor certifications and do not by themselves satisfy technical compliance requirements. Not affiliated with or endorsed by Google, Amazon Web Services, Microsoft, ISC2, Anthropic, or any U.S. government agency.