It doesn't support ad-hoc networking, so although it would be technically possible I imagine the only way for a third-party to implement it would be to record & store on one; download & play on the other.

That said, I think the bigger problem would be triggering it to playback the recording - unless I'm missing something, they just respond to spoken commands.

Which would mean at best you could use it like a voicemail service: "Alexa, play me any messages sent from downstairs".