With the popularity of AI agents & the recent introduction of the MCP by Anthropic. Could we imagine a future with multiple agents with different specialties out there connected through a single protocol & a ‘search engine’ agent that would select an optimal chain of agents for a specific user task?