Table of contents
You’ve probably heard the buzz about Model Context Protocol (MCP) and wondered if it’s just another tech fad. Some developers are saying you don’t need MCP – just let your LLM write API requests directly. But here’s the thing: they’re missing the bigger picture entirely.
Why MCP isn’t just another API wrapper
Let’s clear something up right away: MCP and APIs aren’t competitors. They’re complementary layers that work together. Think of it like this – most MCP servers actually wrap existing APIs under the hood.
The real problem isn’t what APIs can do. It’s how AI models discover, understand, and use them effectively. And that’s where things get messy with traditional approaches.
Here’s what actually happens when you let an LLM write API requests directly:
- Hallucinated paths: The AI makes up endpoints that don’t exist
- Wrong parameters: It guesses at field names and data types
- No error handling: When things break, they break spectacularly
With MCP, the LLM picks which tool to use, then wrapped code executes deterministically. You can test inputs, sanitise data, and handle errors in actual code instead of hoping the AI formats requests correctly. That’s huge for production safety.
The architecture that actually makes sense
The flow looks like this: AI Agent → MCP Client → MCP Servers → REST API → Service
MCP acts as an abstraction layer on top of REST, SOAP, and GraphQL APIs. It’s like having a service layer in your application that wraps API calls. Your database already has a clear query language, but your service layer provides business-logic-appropriate operations.
That’s exactly what MCP does – it provides “semantic APIs” rather than just wrapping existing APIs.
Traditional APIs vs MCP: the real differences
| Aspect | Traditional APIs (REST/GraphQL) | Model Context Protocol (MCP) |
|---|---|---|
| Purpose | Human developers consume endpoints | AI models discover and use tools |
| Discovery | Read documentation, trial and error | Automatic tool discovery and schema |
| Standardisation | Multiple formats (REST, GraphQL, SOAP) | Unified protocol for AI interaction |
| Adaptability | Manual integration for each API | Standardised way for any model to talk to any API |
| Integration | Custom code for each endpoint | Consistent interface across all services |
| Error handling | Depends on implementation | Built-in safety and validation |
| Context management | No context awareness | Provides sufficient context without bloating LLM window |
| Production safety | Manual testing and validation | Deterministic execution with wrapped code |
Why this matters for your marketing stack
The biggest challenge isn’t technical – it’s practical. Most APIs don’t have well-documented OpenAPI specifications that LLMs can actually read and understand. Even when they do, you still need to solve two critical problems:
- How does the client determine when an LLM wants to make a request?
- How do you provide the API schema in a way the LLM can easily structure its responses?
That’s MCP in a nutshell.
MCP standardises how LLMs are expected to call APIs, so any model can talk to any API if it has the right MCP implementation. It provides enough context to prevent LLMs from calling the wrong tool, but stays concise enough to avoid bloating the LLM’s context window.
The bottom line
MCP isn’t about replacing APIs – it’s about making them actually usable for AI systems. When you’re building marketing automation that needs to pull data from your CRM, push to your email platform, and update your analytics dashboard, you want reliability.
You want your AI to pick the right tool and execute it correctly every time. Not guess at API endpoints and hope for the best.
The future of marketing automation isn’t about choosing between MCP and APIs. It’s about using them together to build systems that actually work when your AI agents need to get things done.
We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.
Laura Perrott, Colt Technology Services
Growth Method is the GrowthOS built for marketing teams focused on pipeline — not projects. Book a call to see how it can transform your results.
Additional resources
Want to dive deeper? Check out these resources: