Is FAPI 2.0 being used in Agentic AI flows (MCP?)

[gffletch]

I raised this question in a FAPI working group call a few weeks ago and had a reminder to add an issue once the baseline was migrated to github.

My question is around whether we've seen deployments use FAPI to protect access to MCP servers? Or if there is a need for a higher security API profile in the agentic space? If so, is there any guidance the FAPI working group would want to produce?

[dpostnikov]

Thanks, George. We are talking about a system-to-system API call without user involvement, right?

MCP Client <--> MCP Server