An MCP (Model Context Protocol) is an open standard that lets an AI model plug directly into your software, the way USB lets any accessory plug into any computer. Vendors have started assuming buyers can use one: on a recent onboarding call, a founder told Alec Saluga to skip the web app and connect through the MCP with Claude Code.
Key takeaways
An MCP is a universal connector: one standard plug that lets AI tools like Claude Code read data and take actions inside your software.
A software founder told Alec Saluga on an onboarding call to set up the product through its MCP with Claude Code, assuming he could. An MCP server is just assumed now.
The shift in one sentence: a founder telling a customer not to use the web app.
Alec's framing of where work is heading: you become an agent coordinator, with agents working for you through these connections.
Non-technical users are not locked out: plug-and-play tools like Viktor deliver similar connections with no MCP wiring at all.
What is an MCP, in plain English?
An MCP (Model Context Protocol) is an open standard that lets an AI model plug directly into your software. The best analogy is USB: before USB, every device needed its own custom cable; after USB, one standard port fit everything. An MCP is that standard port, except what plugs in is an AI.
Once connected, the AI can see live data in that tool and take real actions there: read your trades in TradingView, configure your new software, pull a report. The company running the software publishes an MCP server, your AI tool connects to it, and from then on you can just ask for what you want in plain English.
Why do vendors assume you know what an MCP is?
Alec Saluga hit this live on a demo call the week before Session 009. Onboarding onto new software for his business, the founder told him: just use the MCP, hook it up to Claude Code, and have it set it up for you. No survey of his technical background first. The expectation was simply that he knew how.
Read what actually happened there: a founder telling a paying customer not to use the web app. The product's own interface is now the fallback, and the AI connection is the front door. Alec's reaction on air: it's crazy that that's where we are, that an MCP server is just assumed.
“that's a founder of a company telling me not to use the web app. Use the MCP to set it up, and just run it out of Claude Code, not even have to go on the website. Which is a huge shift.”
Alec Saluga · 31:06
What can you actually do with an MCP?
Software setup without the software is the headline example. Alec connected his new vendor's MCP to Claude Code and had the AI do the configuration, the task that would otherwise be an afternoon of clicking through an admin panel.
Live data is the other half. When a stream viewer asked about AI for trading on Session 009, Justin Novak's answer was to connect TradingView through its MCP, which the company ships. The same pattern now covers CRMs, payment tools, and accounting software: anywhere an MCP server exists, your AI can work directly inside that system.
The direction of travel is what Alec keeps pointing at: work turning into being an agent coordinator, with your agents working for you. MCPs are the sockets those agents reach through.
What is the difference between an MCP and an API?
An API is a connection built for programmers: to use one, somebody writes code. An MCP is a connection built for AI models: to use one, you tell your AI what you want, and the AI handles the conversation with the software. Same underlying idea, radically different audience.
The practical consequence is who gets to automate. APIs gated automation behind engineering hires. MCPs hand the same power to anyone running Claude Code, Codex, or an OpenClaw, which is exactly why vendors now lead their onboarding with them.
Do you need to learn MCPs to use AI?
No, and Alec said so himself in the same session. Stringing tools together with APIs and MCPs is how the three operators work, and it is not practical for most people. Plug-and-play products fill the gap: Alec's example is Viktor, an AI coworker in Slack with off-the-shelf connections to Stripe, QuickBooks, and HubSpot, and no real technical setup.
Knowing what an MCP is still pays, even if you never wire one. Vendors will keep saying the word on onboarding calls, and the owners who recognize it will keep getting more out of every tool they buy. At AI4NTP (AI for Non-Techy People) that is the literal mission: knowing what is possible is the edge, and the wiring is optional.
“A lot of what the three of us do with OpenClaw, Claude Code, Codex is piecing things and stringing things together with APIs and MCPs, which isn't practical for most people.”
Alec Saluga · 49:21
Frequently asked questions
What does MCP stand for?
MCP stands for Model Context Protocol, an open standard that lets AI models connect directly to software, read its data, and take actions inside it. Think of it as a universal port that lets an AI plug into your tools.
What is the difference between an MCP and an API?
An API is a connection for programmers, used by writing code. An MCP is a connection for AI models, used by asking your AI in plain English. MCPs put the automation power that used to require engineers into the hands of anyone running an AI tool like Claude Code.
What is an MCP server?
The software company's side of the connection. A vendor publishes an MCP server for its product, your AI tool connects to it, and the AI can then work inside that product. On Session 009, Alec Saluga's new vendor assumed he would set up the product exactly this way.
Do I need to be technical to use an MCP?
Increasingly less so. Connecting one through Claude Code is mostly telling the AI to do it, and Alec Saluga notes that full API-and-MCP wiring is not practical for most people anyway. Plug-and-play tools like Viktor deliver similar connections with no setup at all.
Which tools have MCPs?
The list grows weekly. TradingView ships one, which Justin Novak verified live on Session 009, and business software vendors increasingly lead onboarding with theirs. If you use a tool daily, searching its name plus MCP is now a worthwhile habit.
Tools used in this post
Every tool here has its own page with pricing, who used it live, and honest alternatives.
28:24 Alecliterally on the onboarding call he's like, "oh yeah, just use the MCP and hook it up to Claude Code and have it set it up for you." His expectation was just that I knew how to do that.
28:24 Alecit's crazy that that's where we are, that an MCP server is just assumed.
28:24 Alecworking out of a tool like an OpenClaw, like Claude Code, like Codex, is going to be incredibly important as you see work turning into being an agent coordinator and having your agents working for you.
31:06 Alecthat's a founder of a company telling me not to use the web app. Use the MCP to set it up, and just run it out of Claude Code, not even have to go on the website. Which is a huge shift.
24:39 Justinfor connecting TradingView, use their MCP. I just looked into it, they do have an MCP, so highly recommend checking that out.
49:21 AlecA lot of what the three of us do with OpenClaw, Claude Code, Codex is piecing things and stringing things together with APIs and MCPs, which isn't practical for most people.
49:21 Alecthis is a very, very strong tool for someone less inclined to use a Claude Code or a Codex, given that it's just plug and play. There's no real technical setup.
Who wrote this
Every AI4NTP post is written by an operator who was in the room when the work happened.
A former B2B salesman with no technical background who self-taught AI. He builds and deploys AI-driven marketing and websites, and has grown a following of over 15,000 teaching AI adoption.
Founder and host of AI4NTP. He sold his first company from his college dorm room, and as a fractional CMO has helped scale multiple businesses past $50M in ARR.
A designer, developer, and serial entrepreneur writing code since age 10. He has worked with Disney, the Golden Globes, and the AMAs, and now runs a fleet of AI agents doing real work every day.
