I'm researching and working on the same problem.
This is my proposed stack:-
capturing voice on the client
https://github.com/jiehanzheng/Recorderjs/blob/master/README.md
encoding the captured speech so that it can be shared via Node
https://github.com/jpemartins/speex.js
sharing voice data with multiple users
node.js and socket.io
Somebody wrote a THESIS on this. good read but some missing code in the samples
http://pure.ltu.se/portal/files/36285511/LTU-EX-2012-36247736.pdf
relevant question on so. in particular look at answer by sirjamm
How to Call SPEEX Audio Decode/Encode in HTML5 / JavaScript (Without Flash)
If I get it working, will post the github ref here.
hth