// AGNTcon + MCPcon · San Jose · October 2026

Tom's 14-Week Study Program

MCP Architecture · Agent Orchestration · Frameworks · Observability · Security · Synthesis

📅 Jun 29 → Oct 5, 2026 📚 14 weeks structured learning ❓ 420+ practice questions 🏁 100-question final exam
Overall Progress: 0%

Program Overview

Click any module card to jump to it. Green border = quiz passed.

🎯
HOW TO USE THIS PLAN
Study System
Pass threshold: ≥70% for module quizzes & mid-term; ≥75% for the final exam. Retakes are unlimited — all attempts are logged.
  • Use the nav bar or overview cards to jump between modules.
  • Read / watch / listen to all resources before taking that week's quiz.
  • Quizzes unlock at the bottom of each module card — click Take Quiz.
  • All answers are saved in your browser (localStorage) and survive page refreshes.
  • Visit the 📋 Log tab to review every past answer.
  • Hit Export Log to download a plain-text answer file you can search.
  • Resources marked are optional; free alternatives are always listed.
  • Complete all modules + mid-term, then tackle the 100-question Final in Week 14.
🔌
WEEKS 1–2 · JUNE 29 – JULY 12
Module 1: MCP Architecture
Model Context Protocol — spec, transport layers, primitives, security, ecosystem
New Concept — MCP (Model Context Protocol): An open standard that defines how AI models communicate with external tools and data sources. Think of it as USB-C for AI — a universal connector so any LLM can talk to any tool without custom integration code.
New Concept — Host / Client / Server triad: The Host is the application containing the LLM (e.g. Claude Desktop). Each Client manages one connection to one Server. Servers expose Tools, Resources, and Prompts.
📖 Week 1 — Reading
🎬 Week 1 — Watching
📖 Week 2 — Reading (Go Deeper)
🎧 Week 2 — Listening
Before the quiz, answer these cold: What are the three MCP primitives? When do you use stdio vs SSE? What is the role of Host vs Client vs Server? What is capability negotiation?
🧩
WEEKS 3–4 · JULY 13 – JULY 26
Module 2: Agent Orchestration Patterns
ReAct, Plan-and-Execute, Reflexion, Supervisor/Worker, memory types, handoff protocols
New Concept — ReAct: A reasoning pattern where an agent alternates between Reasoning (thinking about what to do) and Acting (calling a tool) in a loop — like an engineer who thinks before touching the keyboard, checks the result, then thinks again.
New Concept — Supervisor/Worker: An orchestration topology where a supervisor agent decomposes goals and delegates subtasks to specialized worker agents — like a project manager coordinating a team of specialists.
📖 Week 3 — Reading
📖 Week 4 — Reading
🎬 Weeks 3–4 — Watching
🎧 Weeks 3–4 — Listening
Before the quiz: Can you walk through a ReAct loop step by step? How does Plan-and-Execute differ? What does Reflexion add? Name four agent memory types. What is A2A and how does it relate to MCP?
⚡
Week 5 — Mid-Term Exam
July 27 · Tests Modules 1 & 2 comprehensively · 50 Questions · Pass: 70%
This is your checkpoint before the advanced material. Review the checklist below before starting.
📝
MID-TERM PREP
Review Checklist
  • Can you explain the MCP Host / Client / Server triad from memory?
  • Can you articulate when to use stdio vs SSE vs HTTP transport?
  • Can you name and describe all three MCP primitives with examples?
  • Can you walk through a full ReAct loop with a concrete example?
  • Can you contrast Plan-and-Execute with ReAct and say when to use each?
  • Can you describe a Supervisor/Worker topology and its tradeoffs?
  • Can you explain what Reflexion adds beyond basic ReAct?
  • Can you explain A2A and how it complements MCP?
  • Can you name and describe the four agent memory types?
  • Can you explain capability negotiation in MCP?
🏗️
WEEKS 6–7 · AUGUST 3 – AUGUST 16
Module 3: Agentic Stack & Frameworks
LangGraph, AutoGen, CrewAI, Mastra, SmolAgents — forming informed opinions on the landscape
New Concept — LangGraph: A graph-based orchestration framework where agent workflows are nodes (operations) and edges (transitions). Stateful, cyclical, deterministic — the production answer from LangChain for complex enterprise agent workflows.
New Concept — CrewAI: A framework where agents get explicit "roles," "goals," and "backstories" — making multi-agent collaboration feel like managing a team of specialists. Higher abstraction, faster to prototype, less control than LangGraph.
📖 Week 6 — Reading
📖 Week 7 — Reading (Opinion Building)
🎬 Weeks 6–7 — Watching
Opinion to form before quiz: Which framework would you recommend for a regulated enterprise and why? What is "production-ready" for an agent framework? What's missing from the current ecosystem?
📊
WEEKS 8–9 · AUGUST 17 – AUGUST 30
Module 4: Observability & Evals
LangSmith, Langfuse, Helicone, OTel tracing, evals, SLA/SLO translation, cost monitoring
New Concept — Agent Tracing: Recording every LLM call, tool invocation, and decision point in an agent run — like APM for AI. Enables debugging, audit, and performance measurement. The ITIL equivalent is an Incident Record.
New Concept — Evals: Systematic quality measurement of agent outputs — running test suites and scoring against criteria. Like unit tests for probabilistic systems. Hallucination rate → Error Rate KPI. Latency P95 → SLA threshold.
📖 Week 8 — Reading
📖 Week 9 — Reading
🎬 Weeks 8–9 — Watching
ITIL translation exercise: Agent Trace = Incident Record · Eval Suite = Test Plan · Token Cost = Service Cost Model · Latency P95 = SLA Threshold · Hallucination Rate = Error Rate KPI · Continuous Eval = CSI cycle. Map these before the quiz.
🔒
WEEKS 10–11 · AUGUST 31 – SEPTEMBER 13
Module 5: Security & Governance
Prompt injection, least-privilege, audit trails, OWASP LLM Top 10, enterprise governance
New Concept — Prompt Injection: Malicious instructions embedded in data the agent reads (web pages, documents, tool outputs) that hijack its actions. The #1 agent security threat — similar to SQL injection but for LLM reasoning.
New Concept — Least-Privilege Tool Scoping: Each agent should only access the tools and data it actually needs. Minimizes blast radius — a compromised agent can only damage what it had access to.
📖 Week 10 — Reading
📖 Week 11 — Reading (Governance)
🎬 Weeks 10–11 — Watching
Your differentiation: Frame security in ITIL terms — Change Management prevents untested agents reaching production; Configuration Management tracks tool access per agent; Incident Management handles agent misbehaviour. Nobody else at the conference will make this connection naturally.
🎙️
WEEKS 12–13 · SEPTEMBER 14 – SEPTEMBER 27
Module 6: Synthesis & Conference Prep
Opinion formation, debate positions, networking strategy, your 90-second narrative
This module is different: No new technical concepts. The goal is synthesizing everything into confident, conversational positions you can hold in the room. Form opinions. Practice your narrative.
⚔️ Week 12 — The Five Debates (Form Your Positions)
  • ⚔️
    MCP vs A2A — Competing or complementary? SELF-DIRECTED
    Read both specs side-by-side. MCP = model→tool layer. A2A = agent→agent layer. Your answer: complementary.
  • ⚔️
    Autonomy vs Human-in-the-Loop — Where's the threshold? SELF-DIRECTED
    Your ITIL background gives you a structured answer: Change Advisory Board model applied to agent actions.
  • ⚔️
    Framework Sprawl — Consolidation coming or permanent fragmentation? SELF-DIRECTED
    Study how the web framework landscape consolidated (jQuery → React) vs stayed fragmented (CSS frameworks). Which trajectory fits agents?
  • ⚔️
    OSS vs Proprietary models in production — Cost, control, capability? SELF-DIRECTED
    You run both local and cloud models. Form a position on when each is appropriate in enterprise.
  • ⚔️
    Who owns agent governance in enterprise? — IT? Legal? AI team? SELF-DIRECTED
    This is your home turf. The answer from an ITIL perspective is clear. Most developers don't think this way.
📖 Week 13 — Conference Intelligence
  • 📄
    Review the full agenda. Identify 5 speakers you want to meet. Look them up on LinkedIn + GitHub beforehand.
  • 📄
    90-Second Intro Draft SELF-DIRECTED
    "I'm Tom — IT Service Delivery background at Ontario Health, spent the last year building and running multi-agent systems. I'm particularly interested in [X] — how are you approaching it?" Practice out loud.
  • 📄
    Check this weekly in September. Know what's gaining momentum in the month before you arrive.
🎧 Weeks 12–13 — Listening
🏁
Week 14 — Final Exam
September 28 · 100 Questions across all six modules · Pass: 75%
Your comprehensive assessment before heading to San Jose.
✅
FINAL WEEK
Conference Readiness Checklist
  • All six module quizzes completed with ≥70%
  • Mid-term completed with ≥70%
  • Final exam completed with ≥75%
  • Can explain MCP primitives cold — no notes
  • Can name 4+ agent frameworks with clear differentiators
  • Have formed positions on all five debate topics
  • Know 5 speakers at the conference by name and their work
  • 90-second intro practised and natural
  • LinkedIn updated with relevant agent project highlights
  • Reviewed GitHub Trending in the week before departure
  • Listened to Latent Space's most recent episode
📋 Answer Log — all quiz attempts saved in browser
No answers logged yet. Complete a quiz to see your results here.
Quiz
--