A free, structured prep for the Claude Certified Architect – Foundations exam (CCAR-F), in short videos of 5 minutes or less. Chapters follow the five exam domains: tools and MCP, prompts and structured output, agents and multi-agent systems, context and reliability, and Claude Code. Every chapter ends with original practice questions, and every video links to the official documentation.
Independent study material, not affiliated with Anthropic.
Aulas do Curso
(99 vídeos)
0 de 99 aulas concluídas
0%
Progresso salvo somente neste navegador.
0
Claude Certified Architect Exam (CCAR-F) Explained in 5 Minutes
Welcome to Software Engineering Unpacked! This video opens our free Claude Certified Architect – Foundations (CCAR-F) pl…
1
How to Study for the Claude Architect Exam (CCAR-F): Practice First
Don't watch everything from start to finish. Here's how this free Claude Certified Architect – Foundations (CCAR-F) play…
2
Claude API vs Agent SDK vs Claude Code vs MCP: What Each One Is For (CCAR-F)
Four names keep coming up in the Claude Certified Architect – Foundations (CCAR-F) exam: the Claude API, the Agent SDK, …
3
Claude Agent SDK Support Agent: the Running Example, 4 Tools and 4 Rules (CCAR-F)
Meet Cogwheel Bikes, the fictional bike-rental shop and support agent we use in every video of this Claude Certified Arc…
4
Your First Claude API Call: System Prompt, Messages, max_tokens (CCAR-F)
One request to the Claude Messages API, field by field: the three required fields, the system prompt, user and assistant…
5
Claude stop_reason Explained: end_turn, tool_use, max_tokens and All 7 Values (CCAR-F)
Every Claude API response says why it stopped. We go through all seven stop_reason values, what each one means and what …
6
Tool Design and MCP: Claude Architect Exam Domain 2 Explained (CCAR-F)
Chapter 1 of the Claude Certified Architect – Foundations prep: Domain 2, Tool Design & MCP Integration, about 18% of th…
7
How Claude Tool Use Works: tool_use, tool_result and the Round Trip (CCAR-F)
Claude never runs your tools: it asks, and your code runs them. We follow one tool call through the Claude API from star…
8
How to Write Tool Descriptions Claude Actually Follows (CCAR-F)
To Claude, a tool is its name, its description and its schema, so the description decides whether the right tool gets pi…
9
How Many Tools Should an AI Agent Have? Split, Merge and Rename (CCAR-F)
Giving an agent more tools often makes it pick worse. On our bike-rental support agent we apply four fixes: scope tools …
10
Claude tool_choice Explained: auto, any, tool or none (CCAR-F)
Claude's tool_choice decides whether Claude may call a tool, must call one, or must not. We cover the four settings, whe…
11
Claude Parallel Tool Calls: Several Tools in One Turn (CCAR-F)
Claude can call several tools in a single response. We cover what that response looks like, how to run the calls, the on…
12
MCP isError and Claude Tool Errors: Errors an Agent Can Act On (CCAR-F)
When a tool fails, Claude's next step depends on what your error says. We cover is_error in the Claude API and isError i…
13
"No Results" vs "Couldn't Check": Never Hide a Tool Failure (CCAR-F)
An empty result and a failed check can look the same to an agent, and that's how customers get told something false. We …
14
What is MCP? Tools, Resources and Prompts Explained (CCAR-F)
What the Model Context Protocol is, who the host, the client and the server are, and the three things an MCP server can …
15
MCP Resources vs Tools: Catalogs to Read, Actions to Take (CCAR-F)
When something should be an MCP resource, when it should be a tool, and where prompts fit. We cover direct resources and…
16
Claude Code MCP Servers: Local, Project and User Scope, .mcp.json Explained (CCAR-F)
How to add MCP servers to Claude Code with claude mcp add, and which scope to pick: local, project (.mcp.json, shared th…
17
Community MCP Server or Your Own? Tool Descriptions That Beat Built-in Tools (CCAR-F)
When to reuse an existing MCP server and when to build your own, where to find servers today, and how to describe your M…
18
Claude Tools and MCP Practice Questions: 3 Scenarios, Answers Explained (CCAR-F)
Three original scenario questions on tool design and MCP for the Claude Certified Architect Foundations exam: which fail…
19
Prompt Engineering and Structured Output: What the Claude Architect Exam Expects (CCAR-F)
Domain 4 of the Claude Certified Architect Foundations exam in plain words: what it covers, its 20% weight, and its six …
20
Explicit Criteria vs "Be Careful": Prompting Claude for Fewer False Positives (CCAR-F)
Why telling Claude to "be careful" or "be more conservative" doesn't cut false alarms, and how explicit criteria do: wha…
21
Few-Shot Prompting with Claude: Examples Aimed at the Hard Cases (CCAR-F)
When detailed instructions still give inconsistent output, a few well-chosen examples fix it. This video shows how to wr…
22
Fewer False Positives in AI Code Review: Severity Criteria and Noisy Checks (CCAR-F)
Why one noisy category makes developers ignore every review comment, and how to fix it: measure precision per category, …
23
How to Reduce Hallucinations in Claude: "I Don't Know", Quotes First, Verify Claims (CCAR-F)
Three habits from Claude's docs that cut made-up answers: give Claude permission to say "I don't know", have it extract …
24
Claude JSON Output: Tool Use vs Strict Tools vs Structured Outputs (CCAR-F)
Three ways to get JSON from Claude, and what each one guarantees: a tool with an input schema (the exam guide's way), st…
25
JSON Schema Design for Claude: Nullable Fields, "unclear" and "other" Enums (CCAR-F)
How to design a JSON schema that stops Claude from making up values: nullable fields for information the source may not …
26
Claude Structured Output Validation: Pydantic, Zod and Retry With the Error (CCAR-F)
A JSON schema guarantees the shape of Claude's output, not the meaning. We catch semantic errors with Pydantic or Zod, s…
27
When a Retry Can't Help: Missing Data, Stated vs Calculated Totals, detected_pattern (CCAR-F)
Retrying with the validation error fixes format slips and wrong fields, but it can't add information the source never ha…
28
Claude Message Batches API Explained: Half the Price, Up to 24 Hours (CCAR-F)
The Message Batches API runs Claude requests asynchronously at 50% of the standard price, with results within 24 hours. …
29
Realtime or Batch? Choosing Claude's Messages API vs Batches API From the Deadline (CCAR-F)
Batch is half the price, but it only promises an answer within 24 hours. We decide between realtime and batch from one t…
30
Claude Prompts and Structured Output Practice Questions: 3 Scenarios Explained (CCAR-F)
Three original scenario questions on prompts and structured output for the Claude Certified Architect Foundations exam: …
31
Agent Patterns: Claude Architect Exam Domain 1 Explained (CCAR-F)
Chapter 3 of the Claude Certified Architect – Foundations prep opens Domain 1, Agentic Architecture & Orchestration, the…
32
AI Workflow vs Agent: The Simple Rule for Choosing (CCAR-F)
What's the difference between an AI workflow and an AI agent, and when is an agent worth it? Using Anthropic's definitio…
33
Prompt Chaining Explained: Split One Big Job into Focused Passes (CCAR-F)
Prompt chaining breaks one big job into a fixed sequence of calls, each working on the output of the one before, with ch…
34
LLM Routing: Send Each Request to the Right Prompt, Model or Agent (CCAR-F)
Routing classifies a request first, then plain code sends it to a follow-up built for that kind: a short prompt, a small…
35
LLM Parallelization: Sectioning vs Voting, Explained (CCAR-F)
Parallelization runs several model calls at the same time and combines their results in code. Sectioning splits a job in…
36
Evaluator-Optimizer Pattern: One Model Writes, Another Grades (CCAR-F)
Ask a model to grade its own work in the same session and it usually approves it. The evaluator-optimizer pattern pairs …
37
Agentic Loop Explained: How a Claude Agent Works, Step by Step (CCAR-F)
Every agent runs one small loop: call Claude, run the tools it asks for, append the results, and call again until the st…
38
How to Stop a Claude Agent Loop: 3 Wrong Ways and the stop_reason Rule (CCAR-F)
Three tempting ways to end an agent loop break in production: reading Claude's words, counting rounds, and checking whet…
39
Single Agent vs Multi-Agent: When the Token Bill Pays Off (CCAR-F)
A team of Claude agents beat a single agent by 90.2% on Anthropic's research eval, but multi-agent systems use about 15 …
40
Gates, Not Prompts: Enforce Business Rules in Code with Agent SDK Hooks (CCAR-F)
A system prompt is guidance, not a guarantee. We build two gates for a small support agent: a PreToolUse hook that denie…
41
Multi-Agent Systems and the Agent SDK: Claude Architect Exam Domain 1, Part 2 (CCAR-F)
Chapter 4 of the Claude Certified Architect – Foundations prep: the second half of Domain 1, Agentic Architecture & Orch…
42
Multi-Agent Coordinator and Subagents with Claude: Hub and Spoke (CCAR-F)
A coordinator with subagents is the core multi-agent pattern on the Claude Certified Architect – Foundations (CCAR-F) ex…
43
Claude Subagents Start Empty: Pass Context Explicitly (CCAR-F)
A Claude subagent starts with a fresh context: the only thing its parent passes in is the prompt of the Agent tool call.…
44
Multi-Agent Research System with Claude: Split, Fill Gaps, Iterate (CCAR-F)
How a multi-agent research system with Claude finds what a single pass misses. We cover why research needs dynamic decom…
45
Claude Agent SDK Explained: What It Does for You (CCAR-F)
The Claude Agent SDK gives you Claude Code's tools, agent loop and context management as a Python or TypeScript library.…
46
Claude Agent SDK Custom Tools: an In-Process MCP Server (CCAR-F)
Give a Claude Agent SDK agent your own tools: define one with @tool (name, description, schema, handler), serve it from …
47
Agent SDK allowedTools Explained: It Doesn't Mean “Only These Tools” (CCAR-F)
In the Claude Agent SDK, allowedTools pre-approves tools; it doesn't limit the agent to them. We cover availability vs p…
48
Claude Agent SDK Subagents: AgentDefinition, the Agent Tool, Parallel Calls (CCAR-F)
How to spawn subagents with the Claude Agent SDK: define them with AgentDefinition (description, prompt, tools, model), …
49
Claude Agent SDK Hooks: PreToolUse and PostToolUse Explained (CCAR-F)
The Claude Agent SDK hook API in one pass: events, matchers and callbacks; the four answers a PreToolUse hook can give (…
50
Claude Agent SDK Sessions: Resume, Fork or Start Fresh? (--resume, fork_session) (CCAR-F)
What an agent session keeps (the conversation, not your files), the three ways back into one, and how to choose. Continu…
51
Agentic Architecture Practice Questions: 3 Claude Architect Scenarios Explained (CCAR-F)
Three original scenario questions on Domain 1 of the Claude Certified Architect Foundations exam, agentic architecture a…
52
Context Management and Reliability: Claude Architect Exam Domain 5 Explained (CCAR-F)
Chapter 5 of the Claude Certified Architect – Foundations prep: Domain 5, Context Management & Reliability. It's only 15…
53
Claude Context Window Explained: Context Rot and Why 1M Tokens Is a Budget (CCAR-F)
What the context window really holds (everything the model sees for one response, its own output included), how a suppor…
54
Lost in the Middle: Where to Put Long Documents and the Question in a Claude Prompt (CCAR-F)
In a long prompt, Claude reliably reads the beginning and the end, but a finding in the middle can be missed. We cover t…
55
Trim Tool Outputs Before They Fill the Claude Context Window (CCAR-F)
A tool result stays in the conversation and goes back to Claude on every turn, so a verbose one costs you again and agai…
56
Just-in-Time Context for AI Agents: Pointers, Not Payloads (CCAR-F)
Loading every document into an agent's prompt costs tokens on every turn and buries what matters. We cover the just-in-t…
57
Claude Code /compact: What Compaction Keeps and What It Loses (CCAR-F)
Compaction lets a long Claude Code or Agent SDK session keep going by summarizing it, but a summary trades detail for ro…
58
Case Facts Block: Keep Amounts, Dates and IDs Exact in Long LLM Chats (CCAR-F)
Summaries blur numbers: after a round or two, $85, a booking ID and 'by Friday' become 'a refund was discussed'. We buil…
59
Long-Running AI Agents: Progress Files, Scratchpads, Git and Manifests (CCAR-F)
Some tasks outlast one context window, and a crash takes everything that lived only in context. We set up a long coding …
60
Multi-Agent Error Handling: Pass Structured Context, Never Swallow Errors (CCAR-F)
When a subagent fails halfway, what it sends up decides whether the coordinator recovers or makes things worse. We follo…
61
When Should an AI Agent Hand Off to a Human? Escalation Triggers (CCAR-F)
Hand off too early and your agent is a waiting room; too late and a customer is stuck with a bot. We go through the thre…
62
LLM Confidence Scores You Can Trust: Calibration and Human Review Routing (CCAR-F)
A confidence score from the model is only useful once you've checked it against answers you already know. We cover field…
63
Stratified Sampling for AI Extraction: Human Review That Scales (CCAR-F)
An overall accuracy number can hide one document type, or one field, that's badly wrong. We check accuracy by document t…
64
Provenance in Multi-Agent AI: Map Every Claim to Its Source, Claude Citations (CCAR-F)
Summaries of summaries lose their sources, and two good figures quietly become one unsourced average. We keep every clai…
65
Context and Reliability Practice Questions: 3 Claude Architect Scenarios Explained (CCAR-F)
Three original scenario questions on Domain 5 of the Claude Certified Architect Foundations exam, context management and…
66
Claude Code Configuration and Workflows: Claude Architect Exam Domain 3 Explained (CCAR-F)
Chapter 6 of the Claude Certified Architect – Foundations prep: Domain 3, Claude Code Configuration & Workflows, about 2…
67
CLAUDE.md Explained: Managed, User, Project and Local Levels in Claude Code (CCAR-F)
Claude Code starts every session with an empty context, and CLAUDE.md is how your instructions carry over. We cover the …
68
CLAUDE.md @imports and /memory: Keep Your Claude Code Instructions Small (CCAR-F)
A long CLAUDE.md costs context and gets followed less reliably. We cover @imports (relative to the importing file, up to…
69
Claude Code Rules: Path-Scoped .claude/rules with Globs, Explained (CCAR-F)
CLAUDE.md loads in every session, so instructions for one part of the codebase cost context everywhere. Path-scoped rule…
70
Claude Code Skills and Slash Commands: SKILL.md, context: fork, allowed-tools (CCAR-F)
A skill packages a workflow in a SKILL.md file that you run with a slash, or that Claude loads when your request matches…
71
How Claude Skills Load: Progressive Disclosure in 3 Levels, and Why a Skill Won't Trigger (CCAR-F)
A Claude skill loads in stages: its name and description always (about 100 tokens), the SKILL.md body only when triggere…
72
CLAUDE.md vs Rules vs Skills: Always Loaded or On Demand? Claude Code Explained (CCAR-F)
CLAUDE.md, path-scoped rules and skills are all markdown instructions for Claude Code, but they load at different times,…
73
Claude Code Built-in Tools: Read, Write, Edit, Bash, Grep, Glob, Which One When (CCAR-F)
Claude Code reads, searches and changes your code through six built-in tools. We map each job to its tool, settle Grep v…
74
Claude Code in a Large Codebase: Explore Subagent, Scratchpads and Less Context (CCAR-F)
Exploring a big codebase can fill Claude Code's context window before the real work starts, and answers get vaguer as it…
75
Claude Code Plan Mode: When to Plan First and When to Skip It (CCAR-F)
Plan mode makes Claude Code research and propose changes before it edits anything, but it isn't free. We cover what plan…
76
Claude Code Best Practices: Examples, Tests First and the Interview Pattern (CCAR-F)
When Claude Code builds something close but not quite right, more prose rarely fixes it. We cover four techniques: concr…
77
Claude Code in CI: -p, --output-format json and --json-schema Explained (CCAR-F)
To use Claude Code in a CI pipeline, your scripts need fields, not paragraphs. We cover the -p flag and exit codes, the …
78
AI Code Review People Trust: Claude Code /code-review, Fresh Reviewers, Only New Issues (CCAR-F)
An AI reviewer only helps if people read what it posts. We cover the four habits behind code review a team trusts: a sep…
79
Claude Code Configuration Practice Questions: 3 Claude Architect Scenarios Explained (CCAR-F)
Three original scenario questions on Domain 3 of the Claude Certified Architect Foundations exam, Claude Code configurat…
80
Claude Architect Exam Scenario: Customer Support Resolution Agent, Mapped (CCAR-F)
The first of the six exam scenarios in the Claude Certified Architect Foundations exam, described in our own words: a su…
81
Claude Architect Exam Scenario: Code Generation with Claude Code, Mapped (CCAR-F)
The second of the six exam scenarios in the Claude Certified Architect Foundations exam, described in our own words: a t…
82
Claude Architect Exam Scenario: Multi-Agent Research System, Mapped (CCAR-F)
Scenario three of the Claude Certified Architect Foundations exam, described in our own words: a coordinator built with …
83
Claude Architect Exam Scenario: Developer Productivity with Claude, Mapped (CCAR-F)
Scenario four of the Claude Certified Architect Foundations exam, described in our own words: a developer helper built w…
84
Claude Architect Exam Scenario: Claude Code for Continuous Integration, Mapped (CCAR-F)
Scenario five of the Claude Certified Architect Foundations exam, described in our own words: Claude Code in a CI pipeli…
85
Claude Architect Exam Scenario: Structured Data Extraction, Mapped (CCAR-F)
Scenario six of the Claude Certified Architect Foundations exam, described in our own words: a system that turns unstruc…
86
Claude Architect Exam Review: 8 Key Distinctions in Agents, Tools and Context (CCAR-F)
Eight pairs that sound alike on the Claude Certified Architect Foundations exam, side by side: workflow vs agent, the st…
87
Claude Code Key Distinctions: CLAUDE.md vs Rules vs Skills, Scopes, Plan Mode, -p (CCAR-F)
Six Claude Code pairs that get confused on the Claude Certified Architect Foundations exam, side by side: CLAUDE.md vs a…
88
Prompts and Structured Output: 6 Key Distinctions for the Claude Architect Exam (CCAR-F)
Six pairs from prompt engineering and structured output that get confused on the Claude Certified Architect Foundations …
89
How to Reason Through Claude Architect Scenario Questions: A Four-Step Routine (CCAR-F)
A four-step routine for scenario questions on the Claude Certified Architect Foundations exam: find the failure and what…
90
Claude Architect Exam Day: The Online Proctoring Checklist (OnVUE) (CCAR-F)
What to check before you take the Claude Certified Architect Foundations exam online, written only from the exam guide a…
91
LLM Tokens and the Context Window Explained: How Claude Reads Your Prompt (CCAR-F Bonus)
What a token is, how Claude writes its answer one token at a time, and what the context window really holds: the system …
92
Why the Same Prompt Gives Different Answers: LLM Temperature and Sampling Explained (CCAR-F Bonus)
Why Claude can answer the same prompt two different ways: each next token is sampled from a list of likely options. We c…
93
XML Tags and Roles in Claude Prompts: Structure a Prompt Claude Can Parse (CCAR-F Bonus)
How to structure a Claude prompt so nothing gets mixed up: one XML tag per kind of content, consistent names, nesting, l…
94
Build an MCP Server in Python in 5 Minutes: One Tool, One Resource, Claude Code (CCAR-F Bonus)
Build a small MCP server with the official Python SDK v2: one tool and one resource for a bike-rental support agent. We …
95
MCP Transports Explained: stdio vs Streamable HTTP, Local vs Remote MCP Servers (CCAR-F Bonus)
MCP has two standard transports. stdio runs a local server as a subprocess and talks over standard input and output; Str…
96
Claude Code Custom Subagents: .claude/agents, Tools and How Claude Delegates (CCAR-F Bonus)
How to write your own subagents for Claude Code: one markdown file with a name, a description and only the tools it need…
97
Claude Code Permission Modes: default, acceptEdits, plan, auto, dontAsk, bypass (CCAR-F Bonus)
Claude Code has six permission modes, and each one decides what runs without asking you. We cover what each mode allows,…
98
LLM Evals Explained: Build a Test Set Before Every Prompt or Model Change (CCAR-F Bonus)
How to know whether a prompt tweak or a model upgrade made your Claude agent better or broke it. We define measurable su…