I wonder how would we be able to manage thousands of "apps" or "calls" in a way that users can have more control on how, which and when MCP servers are used, and with the right permissions
I would argue that due to the way MCP servers/tools are added to calls, there will be a pre-step that will figure out which MCPs are even relevant for a request prior to executing it.
Yeah. The clients are responsible for doing this right now but I could also imagine an MCP wrapper helping as it gets more complex - or maybe single-focused clients emerge that do a better job for a limited number of MCPs and tasks.
Nice post. A more pessimist view is that MCPs become Amazon Alexa plugins- where you are dependent on an inconsistent Alexa to figure out which plugin to use and limited to only a few.
I wonder how would we be able to manage thousands of "apps" or "calls" in a way that users can have more control on how, which and when MCP servers are used, and with the right permissions
A great question. The ecosystem will be relying on the MCP client for this right now, but I agree it is unclear how this scales.
I would argue that due to the way MCP servers/tools are added to calls, there will be a pre-step that will figure out which MCPs are even relevant for a request prior to executing it.
Yeah. The clients are responsible for doing this right now but I could also imagine an MCP wrapper helping as it gets more complex - or maybe single-focused clients emerge that do a better job for a limited number of MCPs and tasks.
Nice post. A more pessimist view is that MCPs become Amazon Alexa plugins- where you are dependent on an inconsistent Alexa to figure out which plugin to use and limited to only a few.