I’ve been thinking about this a bit recently, and todays ‘Atlas’ announcement prompted me to post.
Quite a lot of the tasks I would like to have an internet-accessing ‘agent’ require 2FA. And at the moment, that basically makes the agents useless as by the time they get to the stage where the 2FA prompt hits them and I need to intervene - it’s only another click or two for me to have accomplished the task in the first place.
In the current world of MCP/A2A/etc - it would be interesting (and possibly quite advantageous to who-ever first got market share) for some sort of agentic 2FA protocol to be developed. I’ve personally no idea how (no need to thank me!) - but having spoken to quite a few people ‘in the enterprise’ - it’s a biggie.
Anyway - just wanted to put it out there and see if anyone else had some thoughts, or if there was a standard/project I was unaware of. I know - possibly - passkeys could come into play. But I’m usually met with blank stares when I mention such things to enterprise vendors.