MCP Squared

What Inspired Us

We love the power of agentic coding, but actually using MCPs in vibe-driven tools like Cursor or Windsurf felt… brutal. Setup was confusing, discovery was a black box, and we were constantly wondering if we were doing it right. We wanted something that felt like casting a spell—not configuring a server.

What We Learned

We discovered that even technically advanced devs avoid MCPs because of friction, not capability. The gap isn’t intelligence—it’s interface. We also learned that the real opportunity lies in making the agent ecosystem composable, discoverable, and plug-and-play.

How We Built It

We built a CLI and in-editor experience that lets devs run a single prompt. This triggers an analysis of their project (language, structure, intent), selects the best-fit MCPs, and auto-installs and configures them. Behind the scenes, we created a lightweight rule engine, plugin registry, and prompt-to-setup pipeline. All of it runs seamlessly inside their existing dev environment.

Challenges We Faced

• Mapping fuzzy developer intent to specific MCPs
• Normalizing config and install flows across rapidly evolving MCP APIs
• Making it all feel fast, smooth, and invisible
• Creating a monetization model that benefits both devs and providers

The Outcome

MCP² makes using agents feel like cheating. One prompt. No YAML. Full power. We turned frustration into flow—and created a gateway for the next wave of AI-native development.

Github: https://github.com/mcpsquared

Built With

• amazon-web-services
• andi
• anthropic
• cursor
• litellm
• porter
• python