It is amazing how many new standards, technologies and frameworks are entering the IT space lately. Coping up with all the new technologies on the horizon becomes almost impossible. Recently, I have been focusing my spare time to the needs of AI models and Agents from the standpoint of integration and a coherent and secure access to corporate information. This is, all in all, one of my fields of specialization and mostly the bread and butter of what my team and I do all day, architect solutions, perform integrations, implement loosely coupled/well encapsulated IT ecosystems that can easily scale securely to the needs of the organization I serve.

Throughout my career, I have been through multiple changes and hype cycles in the industry, and I have enough experience (and a white beard) to understand that not all the technologies currently released on the market will become a standard. Some of them will fail and enter the oblivion of missed technologies while some others will become a dominant pattern in the industry.  I remember the period where I felt crazy to ignore and avoid CORBA, massive J2EE servers, DCOM, EJB etc. just to mention some of them. I think I have skipped and avoided those big monsters by instinct. Luckily, I did it, I do not regret a thing in doing so.

I still remember when a bunch of skeptical colleagues, basing their careers on web services and SOAP, looked at us very strangely, like we were aliens or kids playing in the backyard. They did this because they were the best players in the market while we were “playing” and pushing for REST over HTTPS and modern API gateways, stateless architectures, cloud native solutions and commoditized components. Eventually we became a standard, we are now entirely REST based, thousands of endpoints, millions of API calls, a dominant pattern.

The reason why I am thinking about all of this is simple. Which standard will become dominant in the AI agents, AI gateways, in the overall models’ entropy we see right now?

I don’t have a crystal ball but my bet (and efforts) this time goes on MCP.  The Model Context Protocol looks to me to be very promising, scalable and functioning technology on how applications can provide context to LLMs. https://modelcontextprotocol.io/introduction

Anthropic released the MCP protocol opensource, you can access it here: https://github.com/modelcontextprotocol

I believe that the future of integration passes through this technology, organizations need to integrate frontier models with internal applications, SAAS.  MCP enables all of this, enables gateways architectures where models and solutions can mesh up to build sophisticated ecosystems.

I believe the next few months will be pivotal, crucial for this market spaces and there will be a spike in demand of such skills, better be ready.