What MCP is, and why it makes AI part of your workflow

For most people, AI is a separate tab. You copy something in, ask a question, and paste the answer back wherever it belongs. That's useful, but you're still the one carrying everything back and forth.

MCP, the Model Context Protocol, removes that step. It's an open standard, introduced by Anthropic in 2024 and now supported by most major AI tools, that lets an AI assistant connect directly to other software. With an MCP server in place, the assistant can look things up in that software and make changes there, the same way you would.

My setup

I use Things 3 on my Mac as my task manager, with Claude as my AI assistant, connected through an MCP server. Claude can read my lists, projects, and notes, and it can add, edit, and file tasks directly. No copying and pasting, no switching windows.

The first thing I did with it

I had Claude go through my entire project list and tag every task that AI could complete or meaningfully help with. For each one, it added a short note saying how AI could help: drafting something, researching a question, or comparing options.

That took a few minutes, and the value has kept paying off since:

1. I have an AI-ready queue. I can filter to just the tagged tasks and pick off the ones where a first draft or some research gets me most of the way.
2. I'm not hunting for where AI fits. When I sit down to a tagged task, the note already says what AI can do with it, so I hand it over and get a first pass or the right clarifying questions back.

That setup has grown since. Agents now go through my list twice a day, and I wrote about what they do there.

Why this matters for a business

Notice what that first step was: go through all the work, mark where AI could help, and write down how. That's a small version of a workflow audit, the same exercise I do with a business, except there it covers how work moves through the company, and each idea gets ranked by the hours and dollars it would save.

And MCP, or a direct connection like it, is what turns those ideas into something that runs. Some examples of what connecting AI to a business's own tools makes possible:

• Your CRM: "Which leads from last month haven't heard back from us?" answered from the real data, with follow-up drafts ready to send.
• Your help desk: a new ticket arrives already tagged by topic and urgency, with the customer's recent orders pulled up next to it.
• Your shared drive: "What did we charge the last three customers for a job like this?" answered from your past quotes, not from memory.
• Your calendar and scheduling software: an inquiry gets a reply with real open time slots, instead of a round of back-and-forth.
• Your accounting software: invoices that don't match their purchase orders get flagged for a person before they're paid.

None of these replace anyone. They remove the copying, searching, and retyping that sit between a question and its answer.

If you haven't looked at MCP yet, it's worth understanding. It's what makes AI feel like part of how you work, rather than a separate tab you switch to.