Why engineering and product teams turn cloud notetakers off

Not privacy theatre — concrete problems that show up in a week of meetings.

A bot in the customer call

Discovery calls are the worst place for an unexplained attendee. Some customers ask you to remove it; others quietly say less. JesRecap records the audio your computer plays, so the participant list stays exactly as your customer expects.

Your roadmap on someone else’s stack

Incident timelines, unreleased features, security discussions and customer names all end up in a meeting transcript. Cloud notetakers make that another SaaS vendor to review, and another breach that would be your incident too.

Locked into one AI

Most assistants bundle a fixed model and a fixed prompt. Teams that have already standardised on a particular AI end up paying twice and getting summaries in someone else’s format.

How JesRecap fits the week

Standups and planning

Record the call, get a transcript locally, and pull out decisions afterwards. The optional live transcript is handy when you are chairing and typing at the same time.

User interviews

Dual-track capture keeps your questions and the participant’s answers on separate WAV files, so a research transcript is clean and quotable. Nothing joins the call to make the participant self-conscious.

Incident reviews

During an incident nobody has capacity to take minutes. Record the bridge, then use the transcript to build the timeline afterwards — and keep it on the machine rather than in a third-party tool.

Meeting actions, local or in your own AI

One click gives you a summary, action items, or a custom prompt you write. By default it copies the prepared prompt with the transcript included, to paste into whichever AI the team uses; flip on AI notes on this device and a local model writes the note instead. There is no cloud AI in the app, and it never sends the transcript anywhere itself.

Formats developers can use

Markdown for the wiki or a pull-request description, plain text for a paste, a JSON archive if you want to script something, SRT/VTT for a recorded demo, and the WAV audio — mixed, microphone-only or system-audio-only.

Local searchable history

A searchable library of past meetings on disk, backed by WAV files and SQLite. Both tracks are retained, so you can retranscribe an old interview with a better model instead of losing the recording to a summary.

What stays local

Audio, transcripts and metadata stay on the device. Transcription is a Whisper-family model running on your own CPU or GPU, downloaded once from models.jesrecap.com and reused offline. Speaker labels come from a local diarization model (a one-time 54 MB download). The only network traffic is those downloads and licence activation or checkout — no accounts, no telemetry.

Worth being explicit: there are no integrations, no calendar sync and no team workspace. JesRecap is a local recorder with good exports, and the boundary is the feature.

Record with consent

Say at the top of the call that you are recording, especially with customers and research participants, and check whether your own policy requires written consent for interviews you plan to quote. Local capture means no one sees a bot — which makes announcing it your responsibility rather than the tool’s.

Questions

Which meeting platforms does it work with?

All of them, because it does not integrate with any. JesRecap records the audio your computer already plays via Core Audio process taps on macOS or WASAPI loopback on Windows, so Zoom, Teams, Meet, Webex, Slack huddles or a browser tab all work the same way.

How do summaries work if there is no cloud AI?

Two ways. By default, meeting actions copy a ready-made prompt with the transcript included to your clipboard, and you paste it into whichever AI your team uses. Or enable AI notes on this device and a local model writes the note in the app — a one-time model download of roughly 1.3 GB, then nothing over the network. Either way, model choice, cost and data policy stay on your side.

Is there an API or a Slack integration?

No. There are no integrations, plugins or calendar sync. What you get instead are exports: Markdown, plain text, a JSON archive, SRT/VTT captions and the WAV audio.

Does it work offline?

Yes, after the one-time speech model download. Recording and transcription then run with no network at all.

What does it cost per seat?

From €34.95 once per person, with no subscription. Volume licensing for a larger team is available on request via the enterprise page.

Own the recording, choose the AI

macOS 14.2+ and Windows 10/11. from €34.95 once, no subscription. The 48-hour free trial is coming soon.