A lot of companies are adding "MCP's" to their services.
It means "Model Context Protocol" but really it should just be called a "connector."
It means you can use ChatGPT or whatever AI to ask questions about all your data inside a service:
For example the email company Kit (which I'm using to send out this email right now) has an MCP, so I connected it to ChatGPT and now I can ask questions and it'll give me the answer, like this:
And it made me a helpful chart....which I INSTANTLY saw some mega dips in open rate.
I then promoted it to ask why these major dips happened, and it found that it was from including certain types of links that TANKED email deliverability (for example YouTube shortened links vs full links):
Another MCP I've been using with ChatGPT is Ahrefs (popular keyword research tool I use all the time) and now instead of sifting through complex reports, I can just have ChatGPT do all the work, like this:
Then it pretty much told me what to title and slug my post:
Anyways, this MCP stuff is exciting because more and more companies are allowing you to connect to their services with your AI, so you can do stuff in it quickly.
Imaging instead of logging into Google Calendar you can just tell ChatGPT "Schedule a meeting with John at 11am on Friday about the product" and it'll just do it for you.
That diagram says it all: your AI app is in the middle, and everything else hangs off one simple pipe called MCP. Instead of wiring Slack to Gmail to GitHub to your database with a rat’s nest of custom code, you talk MCP once and plug into anything that speaks it. Stop hand‑crafting one‑off integrations and start treating tools like interchangeable Lego bricks.
What the Diagram Is Really Showing
The AI app in the center only understands MCP. Databases, web APIs, GitHub, Gmail, Slack, even the local filesystem sit around it as MCP endpoints. Each tool just exposes a small MCP interface, and the AI app can read, write, and orchestrate across all of them without learning a new SDK every time.
Why This Beats Custom Integrations
- One protocol instead of seven different vendor SDKs and auth flows
- Swap tools in and out without rewriting your whole AI app
- Ship features faster because you add MCP endpoints, not bespoke glue code
- Keep your AI logic focused on workflows, not plumbing
How Teams Could Use MCP Instead
A SaaS startup could let its AI assistant read GitHub issues, pull customer data from the database, and post status updates to Slack using the same MCP interface for every tool.
An analytics platform could let an AI agent pull CSVs from local storage, enrich them via web APIs, and email reports through Gmail without writing separate integrations for each step.
