Yesterday’s quote revision reminded me how painful copy-paste can be
At 9 p.m. last Wednesday, I was revising a client’s quote in my small Hangzhou studio. The AI had drafted the copy, but I still had to open the spreadsheet manually, find the costs, then paste the data into another window. By the second quote, I was getting frustrated. Then I realized the problem wasn’t that the AI was weak; it was that the tools weren’t connected. And don’t rush to blame yourself for not knowing how to configure it—I got stuck on this step for an afternoon myself.
What is MCP? It’s the adapter between AI tools
MCP is a standard that lets AI tools connect to files, databases, and business software. Once it’s set up, I can tell an AI, “Look up this client’s orders and generate a quote,” and it can retrieve the information within the authorized scope, without me shuttling data between windows.
Some people are already using it. I know Lin Yue, an independent developer. Last year, she took on enterprise automation projects in Chengdu and used MCP to connect her AI assistant with client spreadsheets, reducing repetitive lookups. She also reminded me to start with one low-risk data source the first time, rather than jumping straight into payments or sensitive client data.
Reproducing it today: nearly ¥0, roughly 30 minutes
Money: A free option should get me started, with an estimated cost of ¥0. Time: The minimum version takes about 30 minutes. Technical barrier: I need to enter the connection details as instructed, but I don’t need to write code.
Where do I start? Open the “Connectors / Extensions / MCP” settings in the AI tool I already use, select “Add connection,” and choose an official or trusted sample data source. Once that works, I can switch to my own files. For the first round, I’d put in test data only.
My earlier mistake was dropping a live client spreadsheet into the setup without carefully checking the authorization scope. Luckily, I caught it early. Before configuring anything, I confirm where the data is stored, what permissions it grants, and how to revoke access.
I’d decide based on where I’m at
If I’m just starting out: I’d skip MCP for now. Manual copying may feel clumsy, but it’s the most reliable option while I’m still proving out my quote workflow.
If I already have one or two clients: I’d connect one data source I use regularly and turn “look up information—generate content—human approval” into a repeatable workflow.
If I’m scaling: I’d consider adding more tools and automated tasks while keeping human review and an operation log. Not everyone needs this setup. There’s no urgency to try it now; getting client delivery right still comes first.