Before you configure anything, you need a concrete picture of where this course takes you. Not abstractions — real examples of what changes when you move from Level 2 to Levels 3 and 4.

Quick Reminder: The 5 Levels

Level	Name	Key Characteristic
1	Chatbot	You ask, it answers. No memory, no action.
2	Assistant with Tools	Works on your files when you ask it to.
3	Single Agent	Gets a goal and figures out the steps autonomously.
4	Multi-Agent Pipeline	Multiple agents chain together — fetch, analyze, format, send.
5	Autonomous Network	Agents self-coordinate and self-correct. Minimal human input.

This course covers Levels 3 and 4. Level 5 is where enterprise AI infrastructure lives — we are not going there.

Level 3 in Practice: The Single Agent

Level 3 means Claude takes a goal and executes the full path to get there — without you specifying every step.

Level 2 version of the same task:

You: “Read reports/ppc-bulk.csv and flag keywords with ACoS over 40%.”
Claude: [reads file, flags keywords]
You: “Now sort them by spend.”
Claude: [re-sorts]
You: “Now calculate suggested bids at 25% ACoS.”
Claude: [calculates]
You: “Save to /analysis.”
Claude: [saves]

Level 3 version:

You: “Analyze this week’s PPC performance and give me an actionable report.”
Claude: [reads file, identifies issues, sorts by priority, calculates bid adjustments, saves formatted report to /analysis — all in one go]

At Level 3, you give Claude the goal. Claude figures out the steps.

Level 4 in Practice: The Multi-Agent Pipeline

Level 4 means the work is split across multiple agents running in sequence, each handling one piece of a larger task.

Example: Weekly business digest

• Agent 1 (Data Fetcher): Queries your Amazon Data Hub — orders, inventory levels, PPC performance for the past 7 days
• Agent 2 (Analyst): Takes the raw data, runs analysis — flags anomalies, calculates week-over-week changes, identifies top and bottom performers
• Agent 3 (Writer): Takes the analysis and writes a clean, formatted digest in plain English
• Agent 4 (Sender): Posts the digest to your Slack channel

None of this requires your input. It runs on a schedule. You wake up Monday morning and it is already in Slack.

That is Level 4: agents working in a relay, each doing what it does best, the whole pipeline running without you.

Real Level 3 and 4 Workflows You Will Build in This Course

• Weekly PPC digest — automated analysis of your advertising performance, delivered to Slack every Sunday
• Inventory reorder alert — scheduled check of your stock levels against velocity; alerts you when reorder is needed before you run out
• Listing quality audit — weekly scan of your active ASINs flagging missing A+ content, incomplete bullets, or SEO gaps
• Monthly profit summary — pulls revenue, COGS, ad spend, and fees from your Amazon data and generates a plain-English P&L summary

What Makes This Possible

Three components make Level 3 and 4 work:

1. MCP servers — give Claude live access to your tools and data without you exporting files (covered in Module 1)
2. Claude Task Scheduler — runs Claude agents on a schedule, without you triggering them (Module 2)
3. Chained prompts — multi-step instructions that link agents in sequence (Module 3)

You already have the context files from Course 1 (CLAUDE.md, MEMORY.md) and the folder structure. Course 2 adds the scheduling layer and the live data connections.