Standup Pulse

Inspiration

After a year of running daily standups, the pattern is always the same. Three people give updates that are mostly noise (rambling sentences, side comments, plan-for-the-day mixed with what-shipped-yesterday), and the person taking notes has to scroll back and untangle it. Then someone says 'just summarize what shipped, what is next, and blockers' and everyone nods, but nobody does it.

Standup Pulse is the smallest possible tool that does exactly that. Paste raw notes, get a three-bullet summary, copy as markdown, done.

What it does

One page. One textarea. One button. The agent reads free-form team standup notes and emits:

• What shipped yesterday (concrete deliverables only)
• What is planned today (committed work, not aspirations)
• Blockers (anything blocking progress)

Click Copy and you get a markdown block ready for Slack, Notion, or a status email.

How I built it with MeDo

A single MeDo Deep Build prompt described the UX (textarea, Process button, output card, Copy button) and a calm blue and gray palette. MeDo first produced a requirements document with a page tree (Input Section, Process Button, Output Section, Copy Button) and an acceptance-criteria checklist, then generated the front-end plus a Supabase Edge Function that calls Gemini 2.5 Flash with a structured-output prompt. One follow-up tightened the empty-state copy. That was the whole conversation.

Plugins and APIs used

• Gemini 2.5 Flash via Supabase Edge Function (MeDo default LLM plug)
• Supabase Edge Function for the model call (no DB, sessionless)
• Clipboard API for the Copy button

No auth, no persistence. Input and response live in the browser only.

Challenges I ran into

Getting Gemini to be strict about the three sections when the input is messy. The first version conflated planned-today and blockers when someone said 'I would like to ship X but Y is blocked'. Tightening the system prompt to treat conditional statements as blockers fixed it. The Copy button also needed a graceful fallback for Clipboard API permission denial.

What I learned

The interesting part of MeDo was how much of the build I did not write. The requirements document was a useful artifact in itself. Reading it before generating caught ambiguity. That is a workflow I want to bring to non-MeDo projects.

What is next

• Persistent history per browser (localStorage) with a load-yesterday button
• Slack integration so the summary can be posted to a channel with one click

Built With

• edge-functions
• gemini
• gemini-2-5-flash
• medo
• no-code
• supabase