Action items · owner · timestamp
Every commit
on the record.
Mac Note Taker reads the diarized transcript and pulls out the moments someone agreed to do something. Owner, task, due, source timestamp. Click the timestamp to hear the original sentence. Export the list to Linear or Notion.
- L
Lina · Redo OG image with new palette before Monday night
“I'll redo the OG image - new palette. Done by Monday.”
▶ 29:33Mon May 30
- M
Marko · Verify Stripe coupon redemption cap = 100
“I'll confirm max_redemptions is set to 100 today.”
▶ 07:21Today
- O
Owen · Run a real $0.79 charge end-to-end to verify webhooks
“I'll do a live $0.79 charge to confirm the full flow.”
▶ 37:42Today
- P
Priya · Lock launch date with marketing + send investor heads-up
“Let's lock Tuesday. I'll ping the seed investors tonight.”
▶ 14:08Tonight
Export →
Each item is four fields.
Nothing wasted, nothing missing. The schema is fixed so exports stay clean across formats.
Owner
The person who took the commitment. Resolved against named speakers + calendar attendees, not Speaker 3.
Task
A short imperative phrasing - 'Redo OG image' rather than 'Lina said she would redo'. Reads like a ticket title.
Due
Best-effort inferred from the transcript - 'by Monday', 'end of week', 'before launch'. If absent, we leave it empty rather than guess.
Source timestamp
The audio time the commitment was made. Click to seek - this is the audit trail, not the AI's interpretation.
One-click export
Land where your team already lives.
Linear
One issue per item. Owner -> assignee by email match. Source timestamp + transcript link land in the description.
Notion
One row per item in a database. Columns: Owner, Task, Due, Meeting, Timestamp. Drop into your weekly review.
Markdown
Plain checkbox list with timestamps. Paste into Linear/GitHub/Obsidian/Apple Notes - any tool that accepts md.
JSON
Strict schema { owner, task, due, timestamp, sourceQuote }. For your own automations and CRM hooks.
The timestamp is load-bearing.
Action items pulled by AI are useful. Action items linked back to the audio that produced them are auditable. When someone disagrees with the task we extracted, you click the timestamp and you both hear the exact sentence again.
01
AI proposes
LLM scans the diarized transcript and writes the owner/task/due triple.
02
We snap to turn
Source timestamp snaps to the speaker turn boundary so playback starts cleanly.
03
You click and verify
One tap on the timestamp seeks audio + scrolls transcript. Truth source restored.
Same extraction, your choice of model.
The action-item extraction prompt runs on whatever LLM you pointed Mac Note Taker at. Local Ollama on a customer NDA call, your OpenAI key on a public retro. One toggle, identical output schema.
Action items in every meeting. No copy-paste.
$149 lifetime · 3 Macs · code FOUNDER for $79.