v0.1.0-beta.8 · Public Beta · Free

Download KAVI.
Nothing Leaves Your Laptop.

Record, transcribe and summarise every meeting entirely on your own machine. No account, no cloud, no GPU. Install it and it works offline from the first minute.

100% Local CPU only — no GPU No account needed

Pick your platform

Release v0.1.0-beta.8 · published 17 September 2026

Release notes & checksums on GitHub →

macOS

macOS 12+ · Apple Silicon

Download .dmg

Apple Silicon · 8.6 MB

M1, M2, M3 and M4 Macs.

Intel Macs are not supported yet — an x86_64 build is on the way.

Linux

Ubuntu 20.04+ · Debian · x86_64

Download .deb

Debian package · 17.1 MB

Install with:

sudo dpkg -i Kavi_0.1.0_amd64.deb

AppImage and .rpm builds are coming.

What you need to run it

KAVI runs the AI models on your CPU. If your laptop can open a browser and a spreadsheet at the same time, it can run KAVI.

Processor

Any 64-bit CPU from the last 5 years. No GPU, no NPU, no dedicated hardware.

Memory

8 GB RAM minimum. 16 GB if you want to keep other heavy apps open while recording.

Disk

About 4 GB free. The app is small; the local AI models are the bulk of it.

Internet

Needed once, to fetch the models on first launch. After that KAVI works fully offline.

First launch, step by step

KAVI is in public beta and the builds are not yet code-signed, so your operating system will warn you the first time. Here is exactly what you will see.

1

Install the download

On Windows, run the installer. On macOS, open the .dmg and drag KAVI into Applications. On Linux, run sudo dpkg -i against the .deb.

2

Get past the unsigned-app warning

This is expected during beta and does not mean anything is wrong with the file.

  • Windows: SmartScreen shows “Windows protected your PC”. Click More info, then Run anyway.
  • macOS: Gatekeeper says the app cannot be verified. Open System Settings → Privacy & Security and click Open Anyway.
  • Linux: no warning, but you may need sudo apt-get install -f to pull dependencies.
3

Let the models download once

On first run KAVI fetches its local AI models, roughly 2–3 GB. This happens once. Everything after that runs offline on your own machine.

4

Grant microphone access and record

Allow microphone access when prompted, then hit record. Transcription, summaries and action items all happen locally. New to KAVI? Start with the getting started guide.

Your recordings never leave this machine

There is no KAVI server to upload to. Recordings, transcripts, summaries and tasks are written to your local disk and processed by AI models running on your own CPU. We collect no telemetry from the app.

Read the privacy policy →

Want a nudge when the next build lands?

Beta releases ship regularly. Join the list and we will tell you what changed — or watch the repo if you would rather follow along on GitHub.

Hit a bug? Got an idea?

This is a beta, and the roadmap is still being shaped by the people using it. Crashes, rough edges, missing features, things that felt confusing — all of it is useful. Tell us what you ran into and what you wish it did.

Send suggestions or a bug report

hello@kavimeetings.com · mentioning your OS and version helps us reproduce it faster.