Back to Projects

Minutely – Smart Meeting Debrief
Meeting debrief workspace: paste a transcript and get an executive summary, owned action items with deadlines and priorities, and a structured view of risks and blockers, plus Kanban and markdown export.
Why I built this
After messy meetings, the hard part is not the conversation; it is turning it into something teams can execute. I built Minutely so a raw transcript becomes leadership-ready output without manual note cleanup.
Tech stack
Next.jsn8nGeminiAIApp RouterTypeScript
How I built it
Minutely is a meeting debrief workspace: you give it a transcript, and it turns the conversation into something you can act on: executive summary, owned action items with deadlines and priorities, and a structured view of risks and blockers.
The interface is built for scanning and sharing: tables with clear badges, optional markdown export, and a Kanban-style board so the same debrief can be viewed as columns (summary, actions by priority, risks).
Behind the scenes, the Next.js app does not run the models itself. It orchestrates the UI and proxies requests to an n8n webhook. The n8n workflow runs three Gemini agents in parallel (actions, risks, and summary), then merges the outputs into JSON for the frontend. Configuration uses N8N_WEBHOOK_URL in .env.local pointing at your n8n webhook.
Stack: Next.js (App Router) for the UI and an API route that forwards transcripts to n8n; n8n for multi-agent orchestration and Gemini calls.
Product gallery
Orchestration, structured output, and triage views. Click any image for a fullscreen preview.
n8n multi-agent flow
A POST from the Next.js API hits the n8n webhook with the transcript. A preprocess step normalises the payload, then three branches run in parallel: a Risks agent, an Actions agent, and a Summary agent, each backed by Google Gemini. Their outputs merge into one object, are assembled for the client, and the workflow responds to the webhook with JSON the UI renders.
Summary & action items
Risks & blockers
Kanban view