Model Context Protocol (MCP): How to Connect AI to Your Business Tools Safely

MCP — Model Context Protocol — is the AI infrastructure that's quickly becoming the key layer underneath almost every serious AI setup. It's a big part of why AI is shifting from something where you copy and paste from one tab to the other, to something that actually can act on your behalf, and is the foundation that makes agentic AI possible.

Last week in part one of this series, we covered the what MCP is and why it's such a big deal. In part two, we get into what MCP actually looks like in practice, from the easy non-technical entry points (no technical knowledge required!) already built into Claude to automation tools like Zapier to Kyle's own more advanced setup that allows you to have plain-English conversation with your business data in 30 minutes. We cover a range of options so you can find the right starting point for where you are right now, and understand how far you can take it from there.

If you are a founder, operator or small business owners who is tired of manually looking at data across all your different systems and tools, this is the episode for you.

What You Will Learn

The easiest way to get started with MCP with no technical knowledge required What a more advanced setup looks like using BigQuery and Claude Why clean data still matters — MCP removes the barrier between you and your data, but it can't fix what is broken underneath The safety rules that apply to every MCP setup What multi-agent systems look like next, and why MCP is the infrastructure that makes them possible

Try Granola If you've ever sat in a meeting, taken what felt like decent notes, and then opened them afterwards and they didn't capture anything, Granola is the tool for you. It runs in the background, captures everything, and turns your notes into something you can actually use. Both Jess and Kyle use it, it plays really nicely with Claude, and it is one of our most highly recommended tools

New users get 100% off their first month using our link: granola.ai?via=early-adoptr

Timestamps: 00:00 What We've Been Up To This Week 04:43 What Is MCP and Why Does It Matter? A Quick Recap 08:58 Why AI Agents Need MCP to Actually Be Useful 11:28 The Easy Wins: MCP Connectors Already Built Into Claude 18:12 Zapier, n8n and Make: The Next Step Up 23:40 The Advanced Setup: Talking to Your Data Warehouse With Claude 26:28 The Problem MCP Solves: Getting Answers Without a Developer 28:47 Asking Your Data Questions in Plain English 32:41 Democratizing Data Analysis for All Businesses 34:33 Garbage In, Garbage Out: Why Clean Data Still Matters 36:31 Having a Real Conversation With Your Data: Memory and Context in Data Conversations 39:35 Pulling From Multiple Systems in a Single Question 42:27 Where MCP Is Heading in the Next 12 Months 47:03 AI News of the Week: What 81,000 Claude Users Actually Want From AI 49:06 AI Gone Wrong: The Importance of Human Oversight 52:34 Wrapping Up for the Week

Get in Touch

Email: [email protected] TikTok: @early_adoptr Instagram: @early_adoptr YouTube: @early_adoptr LinkedIn: @early_adoptr

Resources: Official Anthropic MCP server list: github.com/anthropics/mcp-servers GitHub MCP server (what we used): github.com/github/github-mcp-server BigQuery MCP server options: search Smithery.ai for “bigquery” Zapier MCP (no-code entry point): zapier.com/mcp Smithery.ai — browse and discover MCP servers OWASP MCP Top 10 — security reference: owasp.org/www-project-mcp-top-10 AI News of the Week: https://www.anthropic.com/features/81k-interviews AI Gone Wrong: https://fortune.com/2026/03/18/ai-coding-risks-amazon-agents-enterprise/ AI Gone Wrong: https://techcrunch.com/2026/03/18/meta-is-having-trouble-with-rogue-ai-agents/ Get in touch with Early Adoptr: [email protected]

Follow Us on Socials & Resources:

IG: https://instagram.com/early_adoptr TikTok: https://tiktok.com/@early_adoptr YouTube: https://www.youtube.com/@early_adoptr Substack: https://substack.com/@earlyadoptrpod Hosted on Acast. See acast.com/privac