AI for meetings that remembers what was agreed
A meeting lasts until someone clicks Leave. Whisperer keeps the transcript and then breaks it into topics, decisions and tasks — and every task still points at the sentence it came from.
60 free minutes a month · no bot joins the call
- Decision Launch moves to Thursday From “Product standup”
- Task Ivan — pull the churn numbers by Wednesday From “Product standup”
- Task Maria — update the deck for Thursday From “Product standup”
Before → after
Four minutes of talk and what survives them
Below is a fragment of an ordinary standup. Press “Break it down” and see what comes out. Every result remembers its sentence: click a card and the line it came from lights up.
- MarinaLook, we are not making Tuesday. Copy is not ready and half the screenshots are stale.
- PeterLet us move the launch to Thursday. Two days make the difference.
- MarinaFine, Thursday. But then we need the churn numbers for last month, otherwise the sales conversation goes nowhere.
- IvanI will pull the data, done by Wednesday.
- MarinaMaria, can you update the deck for Thursday?
- MariaYes, I will.
- PeterAnd someone has to take the new wording to legal. Not today though.
Nothing yet. This is exactly what a meeting looks like when nobody writes the follow-up.
An example breakdown. The open question stays a question: a person assigns the owner, not a model.
Limits
Two things worth knowing before the first meeting
Neither is a setting you can flip — both follow from how the product works. If either rules out your case, this is the moment to find out.
A finished recording can be uploaded: the Meetings section takes one file up to 25 MB — mp3, m4a, wav, ogg, flac, aac, plus mp4 and webm, from which the audio track is used. What is missing is bulk upload of a whole archive; and it spends the same minutes a live call does.
Participants are heard through system audio: on macOS that means granting the Screen Recording permission, on Windows no extra permission is needed. A physical meeting room with no online call, where only a laptop microphone is available, is not the case this is built for — the transcript will contain you and no one else.
Address
Work needs a place, not a folder called misc
A processed meeting is useless if its results have nowhere to go. Whisperer first works out which piece of work the conversation belonged to, and only then files anything — on that work’s board, not in whichever project came first.
- Meeting A transcript labelled by speaker, plus a meeting map: topics, decisions, tasks.
- Hub A client, a project, a hire, a release. Meetings find their hub on their own; correct it once and you are not asked about that series again.
- Promise What was said out loud is kept next to the meeting it came from. It becomes a task when you decide to track it.
- Board A card on that hub’s board, with a due date, an owner and the source meeting.
Next morning
What to open the next day
All of it lives in your account and is searchable in plain words — including something said three months ago.
Who said what
The full conversation with speaker labels and timestamps. Copy it out or export as plain text or Markdown.
Topics, decisions, tasks
Built automatically after the session. The Enrich button goes back over anything the conversation covered in half a sentence.
“What did we agree on price?”
Ask in ordinary words, get an answer with the meeting and the moment attached. With no source, the assistant says so rather than guessing.
Questions
What people ask first
Does a bot have to join the call?
No. Whisperer is a window above your screen, not a participant. The invite list does not change, nobody sees a stranger in the room, and there is no workspace admin to negotiate with. It works with Zoom, Google Meet, Microsoft Teams, Telegram, Discord and anything else.
Can I upload a finished recording?
Yes. In the Meetings section you pick the file with a button or drag it in — one at a time, up to 25 MB: mp3, m4a, wav, ogg, flac, aac, plus mp4 and webm, from which the audio is taken. From there it goes exactly like a live call: transcript, meeting map, decisions and tasks. A whole archive cannot be uploaded in one go, and in no-logs mode upload is unavailable — nothing is stored there by design.
What if the meeting is in a room rather than a call?
Whisperer will hear your laptop microphone and nothing else. Participants come in through system audio, and an in-person room has none. Speaker labelling will not work the way it does on a call.
Does a model invent the decisions and tasks?
The meeting map is built from the transcript, and every node shows where it came from. A promise made on a call is kept separately from tasks: it becomes a task when you decide to track it. Open questions stay open — a person assigns the owner.
Where do the tasks end up?
On the board of the hub the meeting belongs to. Until a hub has a board, the suggestion declines honestly rather than filing a task into a random project. Every card carries its source meeting, so it is always clear which conversation produced it.
What about confidential conversations?
Switch on no-logs mode: the transcript and the answers are not stored after the session ends, no meeting map is built, and minutes are consumed as usual. Ordinary sessions stay in your account until you delete them yourself.
What does it cost to try?
The Free plan gives 60 minutes of recognition a month. Start, Pro and Max lift the minute limit; the knowledge base and the advanced prompt modes open on the paid plans.
Try it on one standup
Take the next recurring meeting — the one nobody writes up afterwards. Start Whisperer before it begins, then open the meeting map in the morning and compare it with what you remember.