APIs, MCPs, automation platforms, and how to connect AI to everything.
Let's see how your Claude setup held up in the real world.
Last week Claude talked to you. This week it talks to everything else.
Who built a Project this week? Who tried Dispatch? What automated task saved you the most time? Share with the group — the best automations come from real workflows, not hypothetical ones.
"If you set up one scheduled task and it ran every day this week without you touching it — congratulations, you've automated your first workflow. That's the foundation everything else builds on."
Tonight we break out of the chatbox. Claude stops being a tool you talk to and starts being a tool that lives inside your other tools.
Before we connect anything, you need to understand the language that apps use to talk to each other.
An API (Application Programming Interface) is how software talks to other software. When you check the weather on your phone, your app sends a request to a weather service's API, which sends back the data. You never see the request — but it's happening behind every app interaction.
Think of an API like a waiter at a restaurant. You (the customer) don't walk into the kitchen and start cooking. Instead, you tell the waiter what you want, the waiter takes your order to the kitchen, and the kitchen sends back your food through the waiter. The API is the waiter — it takes your request, delivers it to the system that can fulfill it, and brings back the response. You never need to know how the kitchen works.
"You don't need to understand the kitchen. You just need to know how to talk to the waiter."
You manually copy data from one app, switch to another app, and paste it in. Every. Single. Time.
Your apps talk to each other automatically through APIs — no copying, no pasting, no switching tabs.
That's it. Request → Response. Every API works this way: you ask for something, you get something back. The format is almost always JSON — structured data that machines read easily.
You'll never need to write API code yourself. But understanding this pattern unlocks everything else in tonight's lesson. When we talk about MCPs, they're using APIs under the hood. When we talk about Zapier, it's connecting APIs for you. The concept is simple — the power comes from what you connect.
I'm looking at [tool/service name] and want to understand what their API can do. Explain it to me like I'm a non-technical business person: - What data can I pull from it? - What actions can I trigger? - What would a practical automation look like using this API? Keep it concrete — use my business context: [your role/industry]
Model Context Protocol — the universal adapter that lets Claude plug into your tools directly.
MCP (Model Context Protocol) is an open standard that lets AI models connect directly to external tools and data sources. Instead of copying data into Claude, Claude reaches out and gets it itself — from your Gmail, your calendar, your Notion, your file system, whatever you connect.
Think of it this way: APIs are the language apps speak. MCP is Claude's translator — it lets Claude speak that language fluently, to any app, without you doing the translation manually.
Copy your calendar into Claude. Paste your email. Describe what's in your Notion. Repeat every conversation.
Claude reads your calendar, checks your email, and pulls from Notion directly — in real time, every conversation.
MCP servers run locally on your computer through Claude Desktop. Here's how to connect one:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"/Users/yourname/Documents"
]
},
"google-drive": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-google-drive"]
}
}
}"MCP is the difference between an AI you have to feed information to, and an AI that goes and gets the information itself. Once you connect your tools, you stop being the middleman."
Read, search, and manage files on your computer. The simplest MCP to start with.
Read emails, search your inbox, draft responses — without leaving Claude.
Check your schedule, find conflicts, prep for meetings automatically.
Query your workspace, update pages, search across your knowledge base.
Before you reach for external platforms, meet the automation stack that's already built into Claude.
Most people skip straight to Zapier or Make — but Claude ships with a powerful automation suite that handles the majority of what you'd want to automate. No extra accounts, no extra cost, no extra complexity.
"For most people, Claude's built-in stack handles 90% of what you'd want to automate."
Hand Claude a complex, multi-step task and let it work through each stage autonomously. Research, draft, analyze, refine — all in one session. Think of it as a capable colleague who takes a brief and comes back with finished work.
Set tasks to run on a recurring schedule — daily briefings, weekly reports, monthly audits. Claude executes them automatically and delivers the results to you. Set it once, it runs forever.
Fire off tasks to Claude from anywhere — your phone, a quick text, another app. Dispatch turns Claude into an always-available assistant you can command remotely without opening the full interface.
Claude can see and interact with your screen — clicking buttons, filling forms, navigating apps that don't have APIs. If a human can do it in a browser, Computer Use can automate it.
Every morning at 8am, I want you to: 1. Check my Gmail for any emails from clients received overnight 2. Summarize each email in 2-3 bullet points 3. Flag any that need a response today (and draft a response for each) 4. Check my Google Calendar for today's meetings 5. Prepare a brief for each meeting based on recent email threads with that person Deliver everything as a clean morning briefing I can review in 5 minutes.
Claude's stack is powerful — but it has limits. Here are the five walls you'll hit, and when to reach for external platforms.
Claude's built-in automation handles most workflows. But there are five specific scenarios where you'll need to escalate to a dedicated platform like Zapier, Make, or n8n. We call these the five gaps.
Instant triggers needed — webhook fires, action happens in milliseconds
Must run continuously without sessions timing out or needing re-prompts
Thousands of events per hour — bulk processing that would overwhelm a conversation
Simple data moves with no reasoning required — move file A to folder B, done
Automatic retries, fallbacks, and alerting across hundreds of workflow runs
"If your task doesn't hit one of these five walls, you probably don't need Zapier. Start with Claude's built-in tools and only escalate when you have a specific reason."
When Claude's built-in tools aren't enough, n8n is the platform where Claude can generate complete workflows as importable JSON.
When you hit one of the five gaps, you need an automation platform. But here's what makes n8n different from the rest: Claude can generate complete n8n workflows as JSON that you import directly. No clicking through menus. No manual configuration. Just describe what you want, and Claude builds it.
Fully open-source with a transparent codebase you can inspect and modify.
Run it on your own server for free — or use n8n Cloud's free tier to get started instantly.
Describe a workflow to Claude, get importable JSON. No manual building required.
Drag-and-drop interface to see, tweak, and debug your workflows visually.
Built-in AI nodes that connect directly to Claude and other LLMs.
Connect to Slack, Gmail, Sheets, CRMs, databases, and hundreds more out of the box.
I need an n8n workflow that does the following: - Trigger: [what starts the workflow — e.g., "new row in Google Sheets"] - Steps: [what should happen — e.g., "send the data to Claude for analysis, then post results to Slack"] - Output: [what the end result should be] Generate the complete n8n workflow as importable JSON. Include all node configurations and connections. I'll paste this directly into n8n's import feature.
What about Zapier and Make.com? They're solid alternatives — Zapier has the most integrations (6,000+) and the easiest learning curve, while Make.com has stronger visual branching. But neither lets Claude generate complete importable workflows. For AI-native automation, n8n is the clear choice.
n8n handles triggers and actions. Claude handles reasoning and decisions. Together, they're unstoppable.
Here's where it gets powerful: n8n can call Claude as a step in any workflow. That means your automations don't just move data — they think about it. n8n handles the triggers, scheduling, and delivery. Claude handles analysis, writing, and decisions.
Notice the pattern: n8n handles triggers, data fetching, and delivery. Claude handles analysis, writing, and decisions. Each tool does what it's best at.
I need an n8n workflow that uses Claude for the thinking parts. Here's what I want to automate: - Trigger: [what starts it — e.g., "new email arrives"] - What needs reasoning: [the parts Claude should handle — e.g., "classify the email, draft a response"] - What needs reliability: [the parts n8n should handle — e.g., "send the reply, log to spreadsheet"] Generate the complete n8n workflow as importable JSON. Use the Claude/AI node for the reasoning steps. Include all node configs and connections.
"Stop asking 'should I use Claude or n8n?' The answer is both. n8n is the body, Claude is the brain. Together they build automations that actually think."
Let's find the workflows in your week that are begging to be automated.
This is the hands-on exercise. We're going to audit your week and identify the repetitive tasks that are eating your time. Not everything should be automated — but the things that should be? You're leaving hours on the table.
Help me run an automation audit on my work week. I'm a [your role] at [your company/industry]. Here are the tasks I do repeatedly every week: 1. [task 1 - e.g., "Check email every morning and forward important ones to my team"] 2. [task 2 - e.g., "Update a spreadsheet with weekly sales numbers"] 3. [task 3 - e.g., "Send status update emails to clients every Friday"] 4. [task 4 - e.g., "Schedule social media posts"] 5. [task 5 - e.g., "Prepare meeting agendas from notes"] For each task, tell me: - Can it be automated? (yes/partial/no) - Best tool: Claude, Zapier/Make, or both? - Estimated time saved per week - Difficulty to set up (easy/medium/hard) Rank them by ROI — what should I automate first?
Look at Claude's recommendations. Pick the one that's high time-savings + easy to set up. That's your first automation. Don't try to automate everything at once — start with one win, prove it works, then add more.
Use Claude to help you build the automation. Tell it exactly what you want, what tools you have, and ask it to walk you through step by step. Claude is the best automation consultant you've ever had — and it's available right now.
I want to automate this workflow: [describe the task from your audit] Tools I have access to: - Claude Pro (with MCP) - [Zapier / Make.com / n8n — pick one] - [List the apps involved — e.g., Gmail, Google Sheets, Slack] Build this for me step by step: 1. What's the trigger? 2. What are the actions? 3. Where does Claude's reasoning fit in? 4. What should I test before going live? Assume I'm non-technical. Be specific about where to click and what to configure.
"The best automation is the one you actually set up. Don't overthink it. Pick the most annoying repetitive task you did this week and automate it tonight."
Questions? Anything about APIs, MCPs, or automation platforms that you want to go deeper on? Let's discuss.
Next week is about advanced workflows and scaling. We'll take everything you've built so far — your coach, your projects, your automations — and level them up. Multi-step chains, error handling, and building systems that run reliably without you.
We'll also cover how to audit and improve automations that aren't performing well, and how to think about AI automation as a long-term skill, not a one-time setup.
"Week 1 you built the brain. Week 2 you gave it hands. Week 3 you connected it to everything. Week 4 you make it run on its own."
Real questions from the live session — the stuff that doesn't fit neatly into the lesson.
Run your audit, build your first automation, and connect your first MCP. Come back next week and we'll make it all run on autopilot.