Trust6 is multi-model: you choose who to talk to, exchange by exchange, with the composer’s model selector. Whatever you pick, the same rule applies: only pseudonymized text crosses the network.
Managed models (with your account)
Once signed in to your account, managed models are available right in the selector, with zero configuration. The service’s access key is claimed and held by the app’s core: it never appears on screen and you never paste anything.
Your own keys (BYOK)
You can also plug in your existing accounts under Preferences, Providers and keys:
- Anthropic (Claude)
- OpenAI (GPT)
- OpenRouter (many models behind a single key)
- Ollama (local models, see below)
Each key can be checked with the Test button before use. Keys are stored in your system’s secure keychain, never in an app file.
The fully local option: Ollama
If Ollama runs on your machine with a model installed, Trust6 can talk to it like any other provider. In that case, even the pseudonymized text never leaves your computer: the whole conversation is local. It is the maximum-caution option, at the price of more modest models than the big cloud ones.
What never changes
- The write, analyze, review, send flow is identical for every provider.
- Attached files are extracted locally for every provider; no native file ever leaves, period.
- The network payload stays inspectable byte for byte, whatever the model.