Your team's context, on your own machine.
The desktop app runs the full Pomegranate engine locally — capture, curation, and the context library. Use it fully offline with a local model, or bring your own API key.
macOS and Windows builds aren't available yet — Linux only for now. Want one of those platforms? Get in touch.
# download the installer
$ wget https://downloads.pomegranate.expert/pomegranate-standalone_0.1.4_amd64.deb
# check it against the published checksums
$ wget https://downloads.pomegranate.expert/SHA256SUMS
$ sha256sum -c SHA256SUMS --ignore-missing
pomegranate-standalone_0.1.4_amd64.deb: OK
# install and launch
$ sudo apt install ./pomegranate-standalone_0.1.4_amd64.deb
$ pomegranate-standalone
First run · choose how it thinks
Three ways to power it — you pick on first launch.
Fully local
Runs Qwen on your machine. Nothing leaves your computer — downloads a ~6.2 GB model on first run.
Needs a recent GPU and 16 GB+ RAM
Together.ai key
Same Qwen model, hosted by Together.ai under your own API key. Works on modest hardware.
Needs a Together.ai API key
Anthropic key
Claude under your own Anthropic API key — the strongest option for capture quality.
Needs an Anthropic API key
You can switch providers later in Setup. Every mode downloads a small on-device embedding model (~274 MB) on first run.
What it needs
OS
64-bit Linux — Debian/Ubuntu via .deb, anything else via AppImage.
Memory
8 GB RAM with a hosted key · 16 GB+ RAM and a recent GPU for fully-local mode.
Disk
~1.5 GB for the app and embedder · +6.2 GB if you choose the local model.
Network
Needed for first-run setup and sync. Fully-local mode works offline after that.
Before you install
Accounts are invite-only.
The app signs in with your Pomegranate account, and we're onboarding teams one at a time. If you don't have an account yet, book a call and we'll set your team up.