There's some real fundamental things they say that I also strongly agree with. exposeTo for example seems heniously biased. But I really resonated & was energized about webmcp not specifying the actual communication protocol:
> Despite the name, the spec "does not prescribe the format in which tools are exposed" and says browsers may expose them "via Model Context Protocol, other proprietary 'function calling' methods, or any other way." So this is not really an MCP binding: it is a general mechanism for a page to register typed, callable functions for an external caller, not limited to forms or even to AI,
There's some real fundamental things they say that I also strongly agree with. exposeTo for example seems heniously biased. But I really resonated & was energized about webmcp not specifying the actual communication protocol:
> Despite the name, the spec "does not prescribe the format in which tools are exposed" and says browsers may expose them "via Model Context Protocol, other proprietary 'function calling' methods, or any other way." So this is not really an MCP binding: it is a general mechanism for a page to register typed, callable functions for an external caller, not limited to forms or even to AI,