1 Local-First Architecture
AmyNote follows a local-first design philosophy. This means:
- Your saved recordings remain available in the app on your iPhone or iPad
- Your saved transcripts remain available on your device
- AI-generated summaries are returned to and saved in your local library
- Speaker identifications are saved with your local meeting data
- Your preferences and settings are stored locally
“Local-first” does not mean “local-only.” Real-time transcription, AI analysis, high-accuracy transcription, translation, recovery, and similar connected features can upload audio or text for processing. Processing copies and results may be retained on our cloud infrastructure so we can complete, recover, and deliver those requests.
2 When Network Is Required
AmyNote connects to the internet for transcription, AI, account, subscription, analytics, and service-reliability features. The two content-processing flows are:
Real-time Transcription
Audio is streamed through an encrypted connection to a speech service for conversion to text. Some high-accuracy, recovery, or post-recording workflows also upload audio files for asynchronous processing and may retain a processing copy.
AI Features (LLM)
When AI summaries, Ask AI, translation, or other smart features run, the required transcript text and related context are sent through our backend to selected AI providers. Results are returned to your device and may also be retained with the processing job for reliability and recovery.
💡 Tip: Existing local recordings and notes can be played, read, and exported without sending their contents for new cloud processing. Creating transcriptions, running AI features, signing in, and syncing service status require a network connection.
3 Our Trusted Partners
We carefully select industry-leading providers known for their security and reliability:
Cloudflare
Network security, DDoS protection, and edge computing
Microsoft Azure
Speech-to-text, translation, and AI infrastructure
OpenAI
Speech recognition and AI models for transcription, summaries, and chat
Anthropic (Claude)
Claude models, directly or through approved model-routing infrastructure, for advanced AI analysis
We select established providers and configure their API or enterprise services with security and data-protection controls. Provider availability and routing can vary by feature, region, reliability, and model selection. See our Privacy Policy for more detail.
4 Data Flow Transparency
Recording & Playback
100% offline — no network required
Reading Transcripts
100% offline — stored locally
Exporting Files
100% offline — exports to PDF, TXT, Word
Real-time Transcription
Requires network — audio is encrypted in transit to a speech service
AI Summary & Chat
Requires network — content is cloud processed; results return to the device
5 What We DON'T Do
- We don't sell your meeting content or personal data
- We don't use your meeting content for targeted advertising
- We don't claim TLS transport encryption is end-to-end encryption
- We don't use your data to train AI models
- We don't require an account to use core features
6 Questions?
If you have any questions about our security practices, please contact us at:
Email: [email protected]