CCA-F — Demo Code

Working code demos for the Udemy course Claude Certified Architect — Foundations (CCA-F): The Teaching Course.

Each folder maps one-to-one to a specific lecture in the course. Every demo runs end-to-end against real Anthropic APIs, with a small mock fallback where API cost or determinism matters.

CCA-F/
├── S02/  # Domain 1 — Agentic Architecture & Orchestration
│   ├── L04-Watch-The-Loop-Break/            # Broken agent loop → structured error fix
│   ├── L07-Coordinator-Subagent-Demo/       # Multi-agent coordinator + subagents
│   └── L12-Session-Management/              # Agent SDK continue / resume / fork
├── S03/  # Domain 2 — Tool Design & MCP Integration
│   └── L07-MCP-Server-Demo/                 # MCP server + agent that consumes it
├── S04/  # Domain 3 — Claude Code Configuration & Workflows
│   └── L07-Configure-Claude-Code-Setup/     # Before/after Claude Code project setup
├── S05/  # Domain 4 — Prompt Engineering & Structured Output
│   └── L08-Structured-Extractor-Scenario-6/ # tool_use + strict mode + validation-retry + Batches
└── S06/  # Domain 5 — Context Management & Reliability
    ├── L03-Context-Compaction/              # /compact under the hood, real token counts
    ├── L04-Retry-Behavior-529-vs-429/       # Distinguish org rate limit from platform overload
    └── L06-Optimize-Multi-Turn-Agent/       # Prompt caching cost comparison

Each demo folder contains:

• README.md — what the demo teaches, how to run it, what to look for
• requirements.txt — pinned Python dependencies for the demo
• .env.example — required environment variables
• *.py — the runnable Python source
• Any supporting data files (contracts, benchmarks, etc.)

Prerequisites

• Python 3.10 or newer
• An Anthropic API key
• For the Claude Code demo (S04/L07) and MCP demo (S03/L07): the Claude Code CLI installed via npm install -g @anthropic-ai/claude-code

Setup for any demo

cd S0X/L0X-Demo-Name/
python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env and add your ANTHROPIC_API_KEY
python <demo-file>.py

Each demo's README.md has the specific run command and what to look for in the output.

Every CCA-F exam question is essentially "something is broken — what do you do?" The right answer fixes the cause, not the symptom. Four canonical patterns run through both the exam and these demos:

Every demo in this repo teaches one of these patterns.

Official Anthropic documentation (start here for any technical claim)

• Anthropic documentation home — canonical reference for API, models, and pricing
• Claude API — Messages endpoint — the core API surface these demos exercise
• Tool use guide — schemas, strict mode, forced tool_choice
• Prompt caching — breakpoints, TTL, cost math
• Extended thinking — budget, multi-turn preservation on newer models
• Message Batches API — 50% discount async workload
• Model overview + pricing — check before you trust any hardcoded price in demo code

Claude Code

• Claude Code setup and configuration
• Slash commands — custom .claude/commands/
• Hooks — the deterministic enforcement primitive
• CLAUDE.md hierarchy — user/project/subdirectory scoping

Claude Agent SDK

• Agent SDK overview
• Session management — continue/resume/fork
• Custom tools
• MCP integration

Model Context Protocol (MCP)

• MCP specification
• MCP Python SDK
• Reference server implementations — read real MCP server code

Community study references

• paullarionov/claude-certified-architect — the highest-quality open-source CCA-F study guide. Deep technical detail across all five domains with worked exam questions. Recommended as a companion to this course.

Certification

• Register for the CCA-F exam — $99 USD (free for the first 5,000 Anthropic Partner Network employees)
• Official CCA-F exam guide v0.1 (PDF) — request via the registration page above; this PDF is the authoritative reference for domains, task statements, sample questions, in-scope and out-of-scope topics

Books and long-form reference

• Anthropic's engineering blog — architectural deep-dives on Claude systems
• Building effective agents (Anthropic) — the paper that shaped modern agentic design patterns

All demos default to:

• claude-sonnet-4-6 for standard workloads
• claude-opus-4-6 where reasoning depth matters
• claude-haiku-4-5-20251001 for cost-sensitive high-volume paths

If Anthropic ships new model IDs, update the MODEL constant near the top of each demo file (or set ANTHROPIC_MODEL in your .env).

If you find a bug, an outdated pattern, or a demo that no longer runs against the current Anthropic API:

1. Open an issue on this repo describing the demo, the command you ran, and the observed vs expected output
2. Or DM in the Udemy Q&A — the course author reads every question

Pull requests welcome for fixes and clarifications; behavioral changes to demos should go through issue discussion first so we don't drift from what the course lectures teach.

Course material and demo code © Alex Kaziuka, released under the MIT License for educational use. Feel free to adapt in your own production work.

Companion to the Udemy course Claude Certified Architect — Foundations: The Teaching Course.