Models and privacy
Does Mystilink provide model access?
Yes: Newton LLM, an officially hosted model service available after sign-in and billed against prepaid balance. You can also use BYOK: bring your own OpenAI, Anthropic, Google, or DeepSeek account and API key, and the desktop app calls the vendor directly. Both can coexist. See Newton LLM and Models and vendors.
Are my conversations sent to Mystilink servers?
Metaphysics conversations, chart results, and profile cards are stored in the local .mystilink/ folder and are not synced to your account as a whole database. When you use a cloud model, related records can still enter the inference context.
Routine services that go to the cloud include accounts, activation, and payments. If you choose Newton LLM, the prompts and replies for that request are handled by the Mystilink / Newton platform and its upstream model services (see Privacy and local data). On the BYOK path, inference content goes directly to the vendor you chose and does not pass through the Newton platform.
Where are API keys stored?
BYOK keys are stored on your device (workspace and user settings) and shown masked in the interface. Do not share keys in screen captures. The Newton path does not need, and should not use, your account sign-in credentials as a shareable "API key".
Who processes my inference content?
- BYOK: request content is sent to the servers of the vendor you chose and is governed by that vendor's privacy policy.
- Newton: handled by the Mystilink / Newton platform and its upstream model services, governed by the privacy policy.
For sensitive questions, decide for yourself which path to use.
Can I use it without a cloud account?
Installers are publicly downloadable, but activation needs a valid Credential, which means registering and purchasing (or obtaining a trial license). Newton LLM also requires signing in and having balance; BYOK model charges are paid to the third party on your own.
How to judge local models and fallback
Check the actual address, where the model runs, and the fallback configuration. Using a local-model tool does not automatically mean the model also runs on your machine; with cloud fallback enabled, context may move to the backup service. The diagnostic upload toggle and the inference path are separate controls; see the full privacy guide.