8 GB
RAM minimum
Usable, but local AI and capture compete for memory on heavier workloads.
System requirements
macOS
v1.0 requires macOS 26 (Tahoe) on Apple Silicon. This is the only supported platform for v1.
Minimum
| Component | Requirement |
|---|---|
| Chip | Apple Silicon (M1 or later) |
| RAM | 8 GB |
| Storage | 5 GB free (for models and data) |
| macOS | 26 (Tahoe) |
With 8 GB RAM the laptop will feel constrained when Gemma 4 E2B chat and capture run together. Transcription, OCR, and search remain comfortable.
Recommended
| Component | Requirement |
|---|---|
| Chip | Apple M2, M3, or M4 |
| RAM | 16 GB or more |
| Storage | 20 GB or more free |
| macOS | 26 (Tahoe) |
On-device Gemma 4 E2B chat via MLX Swift is optimised for Apple Silicon.
Storage over time
The SQLite database holds OCR text, transcriptions, and embeddings. Expect roughly 500 MB per month for database growth on normal use, bounded by configurable retention. No screen images or video files are persisted, so disk use tracks text and metadata only. Keep 20 GB free if you want headroom for models and caches.
Permissions
- Screen recording: required.
- Microphone: required only if you enable audio transcription.
- Accessibility: optional. Improves window and context detection on macOS.
For managed devices, confirm these can be granted under your endpoint management policy before deploying widely.