Kind of an interesting twist in IT history: Did it need the advent of AI Agents

Kind of an interesting twist in IT history: Did it need the advent of AI Agents to drive developer friendly, standardized and discoverable interfaces to systems?

Recently Model Context Protocol (MCP) got a lot of attention and traction. While this is good and exciting thing, I was wondering why this is actual needed? In a perfect & developer friendly world, it shouldn’t need another protocol just for agents, does it? AI should easily be able to use the existing interfaces, which have been built for convenience of human developers, no? (To be fair to MCP: it is not a rival API — it is a standardized discovery-and-invocation layer through which AI applications reach tools and context, and it typically sits on top of exactly those existing interfaces.)

In Martin Keen’s video “MCP vs API: Simplifying AI Agent Integration with External Data”[1] you find a very good overview and comparison of MCP and APIs. Highly recommended to watch!

Still interesting to witness that HATEOAS (Hypermedia as the Engine of Application State) and hypermedia types like siren, HAL and others never really got mainstream traction — and while API specifications based on the OpenAPI standard are widely used today, the fuller package of consistently good specs plus sample client implementations shipped as part of API provider deliverables rarely came together. In my experience both business stakeholders and developers rarely saw the value in this. Too much extra effort, not direct business value, …. Eventually ending up with developers taking on the integration efforts required to integrate systems. Same effort spent over and over again. Cost not being accounted for.

Now this seems to change. Capability providers are visibly investing in MCP server implementations to be part of the Agentic AI future. This is good, but maybe we should have done that already in a pre-AI Agent world?

Note that we might run into a cyclic dependency here as Martin also points out in his video, many of today’s MCP servers are just abstractions of existing APIs (though others expose filesystems, databases, or local tools directly), which all the shortcoming above. But here I expect AI systems to be a great support for developers writing their MCP implementations.

Just a curiosity and point of view developed during my morning jog earlier today -what is your take on this?

[1] https://www.youtube.com/watch?v=7j1t3UZA1TY

• AIAgents
• MCP
• API