The internet of agents is a blue-sky paradigm shift. What will it look like?

Search platforms like Perplexity, You.com, and Google‍ and agent platforms like MultiOn and Adept have started to show us a hint. Search is increasingly going to be done for us, rather than by us.

My POV: these agent web lanes need to be both information-rich and action-rich superhighways. Meta tags and robots.txt have already given us a taste of this. End users don't see this content, but search engine bots use them to more deeply understand your website and business.

Information richness refers to the level at which a user can understand the content, context, and intent of a website. In today's websites, information is primarily communicated at the visual level, with the exception of meta tags, etc. The lingua franca of LLMs, however, is text (which includes specs and code). If we optimize for this, the resulting AI agent communication protocols will allow them to prompt each other for deeper understanding, and provide all of the relevant information based on the context of the end user. Today, we have meta tags. Tomorrow, agents will have meta conversations - and all within milliseconds.

Action richness has less precedent on the web today, but is crucial to the "internet of agents". Today's agents have significant difficulty navigating UIs built for human users. HTML, as it stands, is not the best communication language for agents. However: all is not lost! LLMs are very capable of interacting with APIs given proper specs, and in the future, most "APIs" may be expressed and manipulated by natural language. Agent communication and action languages have existed for some time (since at least the '90s with KQML), but who will create the standard for LLM agents on the web: agents.txt?

This paradigm shift is closer than we think, and Coframe's mission is to deliver the greatest possible value it has to offer to brands and individuals.