Most meeting tools receive one mixed audio stream and try to reconstruct who said what afterward. Discord gives us a better starting point: a separate voice stream for each participant.
Kassin o keeps that identity from capture to transcript, then turns the call into searchable notes, decisions, tasks, and answers with source links. The same authorized meeting memory can be queried from Discord or through a read-only MCP connector.
It is free, AGPL open source, and self-hosted. The public demo uses fictional data, so anyone can explore a finished meeting without signing in.
We launch July 16. I m curious: if your team could ask one reliable question across all its past calls, what would it be?
Love that it keeps one audio track per speaker instead of mashing them together. Makes the transcripts way more accurate when people talk over each other, and the MCP integration is a smart move for teams already living in AI tooling.
@metehankertik thats the idea! Thank you for sharing :)
The per-speaker track separation that flows into named transcripts is a thoughtful detail. Most tools give you a blob of text, but tying each line back to its actual voice makes the search and source links genuinely trustworthy instead of just a nice idea.
@potuktunah15112 Thank you! Your feedback it's more than appreciated s2