MCP: The REST for AI?

Before REST transformed web APIs, every integration was a bespoke challenge—each system spoke its own language, forcing developers to build custom solutions from scratch. Today, as Artificial Intelligence becomes ever more integral, connecting AI models to our systems can feel like trying to fit square pegs into round holes.

Enter Model Context Protocol (MCP). This emerging standard promises to streamline AI-to-system communication in much the same way that REST unified web services. MCP provides a universal language for AI, enabling models to access data, invoke tools, and execute actions without the need for countless custom API bridges.

What MCP Brings to the Table

MCP standardises the way AI models interact with external systems, delivering several key advantages:

• Dynamic Discovery: Instead of relying on fixed endpoints, MCP allows AI models to locate and utilise available tools in real time—imagine browsing a digital menu of services.
• Effortless Integration: A single, consistent protocol means that complex, bespoke integrations are replaced with a streamlined process, reducing development overhead and accelerating deployment.
• Interactive Communication: MCP supports continuous, two-way interactions, enabling AI to maintain context and adapt its actions dynamically rather than being limited to simple request-response cycles.

A Firsthand Look at MCP in Action

In my recent experiments with Claude Desktop, I set up an MCP server in minutes and witnessed its potential firsthand. The process was refreshingly straightforward: my AI model was able to query databases, fetch live data, and trigger actions seamlessly. Alongside Claude Desktop, tools like Cursor, Windsurf, and the Cline plugin are already beginning to adopt MCP, proving that this standard is not just theoretical—it’s actively simplifying real-world workflows.

MCP vs REST: A Comparison for the Modern Era

While it’s tempting to call MCP the “REST for AI,” the analogy only partially holds. Here’s how they differ:

• Fixed Endpoints vs. Dynamic Discovery: REST APIs require developers to know endpoints in advance. In contrast, MCP empowers AI models to discover available services on the fly, adapting as new tools emerge.
• Human vs. AI Consumption: REST was designed with human developers in mind, whereas MCP provides machine-readable schemas that let AI interact with systems in a natural, intuitive manner.
• Static vs. Interactive Communication: REST follows a straightforward request-response cycle. MCP, however, supports a more interactive dialogue, allowing AI to adjust and maintain context during ongoing interactions.

Addressing the Security Challenge

No new standard is without its hurdles. MCP’s security framework is still evolving. While REST benefits from mature authentication methods like OAuth, many current MCP implementations rely on user consent and manual token management. This means that deploying MCP—especially in sensitive or enterprise environments—requires extra care until more robust, automated security measures are in place.

Final Thoughts: Standardisation for Future-Proof Systems That Delight

As a (cracked) software engineer, with a deep commitment to outstanding user experience (UX) and developer experience (DevEx), I am passionate about building systems that are both robust and a joy to use. Standardisation is at the heart of creating future-proof tools, and MCP is a clear example of this vision in action. By making our core systems ready for external AI integrations, we pave the way for large language models (LLMs) to interact with our technology through natural language.

This means that soon, non-technical users might effortlessly engage with complex systems simply by speaking or typing in plain language. Imagine asking your system for a status update, scheduling tasks, or retrieving data—without needing to understand the intricacies of API calls. That is the power of standardisation: it enables us to build products that not only function flawlessly but also bring genuine delight to both users and developers.

MCP is not just about a new protocol—it’s about setting the foundation for a future where AI integrations are plug-and-play, and our systems are agile enough to meet evolving technological demands. Let’s keep pushing the boundaries and build a future where technology truly works for everyone.