title
Please take a moment to fill out this form. We will get back to you as soon as possible.
All fields marked with an asterisk (*) are mandatory.
Claude Agentic Systems
Duration
3 Days
Course
CL-300
Available Formats
Classroom Training, Online Training
Course Description
Overview
This Claude Agentic Systems course is the hands-on technical deep dive shared by the Architect and Developer tracks, and it carries most of the program's implementation weight. Over three days, participants build the systems that the CCA-F exam weights most heavily: agentic loops and multi-agent orchestration, Claude Code configuration and CI/CD workflows, MCP servers in code, and prompt engineering for structured output.The course is build-led, but each major build is preceded by the design decision behind it — when to add an agent versus a tool, how to decompose a task, what to make deterministic — so learners leave able to both design and implement. All instruction is hands-on with the Claude API, the Agent SDK, MCP, and Claude Code.
Objectives
Upon completion of the Claude Agentic Systems course, students will be able to:
- Make core agentic design decisions before building — agent versus tool, decomposition strategy, and what to make deterministic
- Implement agentic loops with correct stop_reason handling, tool result management, and termination logic
- Build multi-agent systems using hub-and-spoke coordinator patterns with correct subagent configuration and context passing
- Configure CLAUDE.md hierarchy, custom slash commands, skills, and path-specific rules for team development
- Integrate Claude Code into CI/CD pipelines using non-interactive mode and structured output
- Build MCP servers in Python with effective tool interfaces and structured error responses
- Apply few-shot prompting and tool use with JSON schemas to enforce structured output in production
- Implement validation, retry, and feedback loops for extraction and classification reliability
- Manage context windows across long documents, multi-turn conversations, and multi-agent handoffs
Audience
Solution architects and developers implementing Claude-based agentic systems. Shared course across the Architect and Developer tracks; required for both on the path to certification.
Topics
- Agent versus tool: when a capability belongs in a prompt, a tool, or a separate agent
- Task decomposition: single loop, pipeline, or coordinator-subagent
- What to make deterministic versus model-driven, and why it matters for reliability
- Lab: scope two scenarios into concrete build plans before writing code
- Lifecycle: request, stop_reason inspection, tool execution, result return
- Model-driven decision-making versus pre-configured tool sequences
- Anti-patterns: arbitrary iteration caps and natural language termination
- Lab: implement a correct agentic loop with multi-tool control flow
- Hub-and-spoke coordinator architecture
- Task decomposition, delegation, and result aggregation
- Subagent configuration: the Task tool, allowedTools, and explicit context passing
- Parallel subagent spawning via multiple Task tool calls
- Lab: build a multi-agent research pipeline with coordinator and subagents
- PostToolUse hooks for data normalization and compliance
- Tool call interception for prerequisite enforcement
- Hooks versus prompt instructions: when deterministic guarantees are required
- Lab: implement a compliant multi-step workflow with prerequisite gates
- CLAUDE.md hierarchy: user, project, and directory levels
- @import syntax and .claude/rules/ for modular configuration
- Custom slash commands and skills with frontmatter configuration
- Path-specific rules with glob pattern scoping
- Plan mode versus direct execution
- Lab: configure a multi-level CLAUDE.md for a simulated team repository
- The -p flag for non-interactive pipeline execution
- --output-format json and --json-schema for structured findings
- Session context isolation: independent review instances versus self-review
- Lab: build a CI integration that posts structured code review output
- MCP server structure, lifecycle, and transport mechanisms
- Implementing tool handlers with correct input and output schemas
- Exposing content catalogs as resources to reduce exploratory tool calls
- Structured error responses and the isError flag pattern
- Lab: build and deploy an MCP server for a realistic integration scenario
- Project-level (.mcp.json) versus user-level scoping
- Environment variable expansion for credential management
- Connecting MCP clients to Agent SDK workflows
- Tool distribution across specialized agents and tool_choice configuration
- Explicit categorical criteria versus vague instructions
- Few-shot prompting for consistent format and edge case handling
- Tool use with JSON schemas for guaranteed schema-compliant output
- tool_choice: auto, any, and forced selection
- Schemas with nullable fields and extensible enum patterns
- Retry-with-error-feedback for self-correction
- Distinguishing retryable format errors from unresolvable absent information
- Context window management: persistent data blocks and tool output trimming
- Batch processing and multi-instance review architectures
- Design a coordinator-subagent system for a realistic scenario (e.g., a research-and-synthesis pipeline or a support-resolution agent with escalation)
- Implement the agentic loops, subagent delegation, and MCP-backed tools the design requires
- Add structured-output enforcement, validation, and human-in-the-loop gates where the design calls for them
- Demonstrate the working system and walk through the design decisions behind it
Self-Paced Training Info
Learn at your own pace with anytime, anywhere training
- Same in-demand topics as instructor-led public and private classes.
- Standalone learning or supplemental reinforcement.
- e-Learning content varies by course and technology.
- View the Self-Paced version of this outline and what is included in the SPVC course.
- Learn more about e-Learning
Course Added To Shopping Cart
bla
bla
bla
bla
bla
bla
Self-Paced Training Terms & Conditions
??spvc-wbt-warning??
Exam Terms & Conditions
??exam-warning??
??group-training-form-area??
??how-can-we-help-you-area??
??personalized-form-area??
??request-quote-area??
Sorry, there are no classes that meet your criteria.
Please contact us to schedule a class.
STOP! Before You Leave
Save 0% on this course!
Take advantage of our online-only offer & save 0% on any course !
Promo Code skip0 will be applied to your registration
Nothing yet
Purchase Information
??elearning-coursenumber?? ??coursename??
View Cart
title
Please take a moment to fill out this form. We will get back to you as soon as possible.
All fields marked with an asterisk (*) are mandatory.