Aurora SwytchCode
Link to open source: https://github.com/himanihassija/Swytchcode-Aurora-Himani-Hassija
Aurora — an AI Meeting & Follow-up Assistant, built on Swytchcode
Aurora automates the complete meeting lifecycle end to end: proposing times and agendas, summarizing call transcripts into clear notes with owned action items, and sending personalized follow-up emails grounded in what was actually discussed. Rather than a single prompt trying to do everything, it runs as three specialized agents — a Scheduler, a Summarizer, and a Writer — handing work off to each other, with their reasoning visible at every step.
What sets it apart is that it doesn't just execute tasks once and stop. Aurora keeps a persistent memory of every meeting it touches, so it never repeats itself or re-contacts someone too soon. And when a follow-up goes unanswered, it doesn't just wait — it detects the stalled thread, flags it as "at risk," rewrites a more direct re-engagement message, sends it, and escalates to a human as a safety net. That combination of visible reasoning, memory, and self-initiated recovery is what turns this from simple automation into something closer to real judgment.
All execution runs live through Swytchcode (Google Calendar, Resend, Notion, with Slack as a bonus escalation channel), and the whole pipeline is visible through a real-time dashboard with a built-in chat interface for asking questions about any past meeting.
