--- published: true layout: post title: HTTP APIs and GraphQL, Event-Driven, and Model Control Protocol date: 2025-03-31T09:00:00.000Z tags: - HTTP - Artificial Intelligence - MCP - GraphQL - Event-Driven image: https://kinlane-productions2.s3.amazonaws.com/algorotoscope-master/francis-williams-mining-mountains-2.jpeg --- It is easy to get caught up in any given moment of the API space, and much more difficult to keep yourself elevated above the space. It is lonely up in the tops of the peaks above the API space, and much more exciting down in the city and streets. I am purposefully staying out of AI discussions as I do with most applications that put APIs to work. Applications definitely shape the API conversations, but often tend to distort the API conversation as well. The application of our digital resources and capabilities using APIs should be our North Star, with specific applications shaping and guiding our approach, but not dictating our overall approach to defining, designing, and operating our APIs across many different types of applications. [Kevin Swiber has a post I still need to review](https://www.layered.dev/mcp-the-ultimate-api-consumer-not-the-api-killer), as I trust their approach to APIs, and the specifications we use, but I was thinking more about [something they said last week](https://www.linkedin.com/feed/update/urn:li:activity:7311380579059466240?commentUrn=urn%3Ali%3Acomment%3A%28activity%3A7311380579059466240%2C7311438412354318336%29&replyUrn=urn%3Ali%3Acomment%3A%28activity%3A7311380579059466240%2C7311459951049089024%29&dashCommentUrn=urn%3Ali%3Afsd_comment%3A%287311438412354318336%2Curn%3Ali%3Aactivity%3A7311380579059466240%29&dashReplyUrn=urn%3Ali%3Afsd_comment%3A%287311459951049089024%2Curn%3Ali%3Aactivity%3A7311380579059466240%29), which has helped align my view of MCP and the wider API universe:
“With APIs, you have more opportunity for serendipitous reuse. MCP handles LLM use cases pretty specifically. I don’t suspect we’ll see it replace other styles of APIs. Like GraphQL is becoming in the enterprise, we’ll probably see MCP as additive. You might want MCP-style instead of OpenAPI-style if your only use case is conversational UI. As soon as you need more, you’ll be moving that functionality to an OpenAPI-style API.”
This helps. I was seeing MCP as a straight assault on OpenAPI like RAML was, but positioning as another API pattern layer that is responding to the needs of a specific application of our digital resources and capabilities pencils out for me. In the world of APIs, every meaningful movement forward tends to come with a vendor and investor induced grab for power, which I am alluding to in [my API specification fragmentation and rot narrative](https://apievangelist.com/2025/03/25/api-specification-fragmentation-and-rot/). People building GraphQL solutions genuinely believed their approach to APIs would replace resourceful approaches. People building event-driven solutions genuinely believe that asynchronous would displace synchronous. Their investors 10x believe that and throw coal into the storytelling furnace around their portfolios—-resulting In that heart pumping velocity of the hype train that comes with each wave. Now, AI is special only in that it comes with a lot more money and science fiction and labor delusion than desktop, web, mobile, and device applications mustered. However, Kevin has helped me align artificial intelligence, or specifically MCP with GraphQL and Event-Driven approaches, in that the champions aren’t wrong. There are very meaningful and useful applications of GraphQL, event-driven, and model-driven approaches to producing and consuming APIs. However, the vendor and investor bluster out on the web rarely matches what is actually happening on the ground within the enterprise. Sure there are eager champions of these tools and real-world use cases of these tools within the enterprise, but the alignment of this reality with the expectations, promises, and appetite of the outside startup and venture fueled noise is always mismatched. Time after time I have come across GraphQL implemented when nobody asked for it, as well as situations where GraphQL would have been sensible, but does not get applied because the hype around it scares people off. Time after time I have come across expensive multi-year projects to deliver Kafka when there was a need for event-driven approaches, when a simple cost-effective approach using Webhooks would have made much more sense. We will watch the same thing play out with artificial intelligence, MCP, and a model-driven approach. Except the hangover from this moment will be much more painful than it has been in other application areas. We’ll see lots of model-driven approaches implemented when they aren’t needed, and other situations where they are needed, but not implemented because the hype scared people off. Over time, the API toolbox will continue to settle, with simple pragmatic HTTP continuing to be the center, but a graph, event, model, and other patterns continuing to ebb and flow at the edges. I have added MCP into my API Pulse report. I will be documenting evidence of MCP in the wild, alongside AsyncAPI, GraphQL, and other implementations. I will take some time this week to read more about what Kevin has written on the subject of MCP. I don’t think Kevin is immune from what is happening at this moment, but I do know that this isn’t Kevin’s first rodeo, and they have seen some shit. I will continue to stay back from what is happening and remain safely above the API space. While sometimes I get concerned that too much time at these high altitudes means that I can become out of touch with what is happening, but in some moments (like these), I actually feel like it is actually after up here. I feel like the API narrative is so dominated by the technology sector and the money behind it all, that I am better staying at the higher elevations, but also continuing to explore the realms of mainstream industries like finance, healthcare, mining, shipping, and other spaces that aren’t necessarily riding the hype train, or even remotely close to the technology sectors Las Vegas vision of APIs, I mean AI, I been blockchain.