
I’ve been experimenting with a lot of AI tools lately. Many of those experiments will never see the light of day, but the Model Context Protocol (MCP) feels like a tool that could redefine agent-to-agent and AI-to-data use cases.
MCP is a protocol that enables AI to interact with data at scale, reducing the need for custom integrations. There’s been a lot of excitement around MCP because it allows AI agents to interact with… well, almost anything.
Process
For this experiment, I wanted to use MCP to build a backend for a Next.js dashboard. The frontend design came from Dribbble—I searched for “dashboard,” picked a random design, and dropped a screenshot into Cursor.
Through Cursor’s IDE, I used MCP to set up and connect to a local instance of Supabase. Supabase, an open-source competitor to Firebase, offers a prompt library that makes it easy to spin up a database. Using MCP, I defined the data model and created the necessary tables.
What I Learned
The scope and speed of what’s possible with AI right now is incredible. I was able to go from an idea to a deployed product in just a few days. Having spent 15 years in product design and 4 years in web development, that level of productivity is astonishing to me.
Repeatable AI processes will only get faster, which means production capacity for digital products is no longer a major constraint. The real challenge ahead will be maintaining differentiation—especially when tools can instantly recreate anything they have MCP access to. The future is going to be really interesting.
Perhaps products, companies, collectives, and / or organization will codify prompts into Prompt Libraries. The lifecycle of a prompt library – creation, codification, governance, and continued enhancement will lead to “Prompt Systems” that blend human values with corporate (in the broader sense of the term) values.
Tools:
- Model Context Protocol
- Cursor AI
- Tailwind CSS
- NextJS
- SupaBase
- Github
Source Code: