Mahmut Canga

SaaS applications have been having a good time but what would happen with all these AI agents?

They would turn into SeaS…

Service as Software!

With the raising trend of MCP protocol, I would expect the SaaS API proving to be useful to my agent by providing their MCP endpoint. Instead of SSO, redirecting or embedding the UI to my business use case, my internal UI for the business will yield an agent workflow that could figure out how to use it and achieve the goals.

This obviously challenges a lot of integration patterns and forces us thinking out of the box for distributed systems challenges too.

Simple questions like how my agent will deal with back pressure? How my agent can work around if it is being throttled? Can my agent provide enough tracing to a human operator to help in an emergency?

Other questions like orchestrator vs choreography seem not easily addressed too. Recent agent workflow examples focus on more orchestrator approaches but choreography through the events could be also interesting to test.

Would that also mean we need something like EMCP? Event Model Context Protocol?

Sea is rising! All of these existing patterns for the new paradigm need a test of time and production battles.

This is why it would not feel Software being the service here anymore.

It really feels like we will be Servicing our business as Software…

Ahoy! SeaS… Service as Software!