Choosing your AI setup
Compare Saymail Cloud credits, bringing your own LLM key, and running a local model.
Saymail’s AI features — email classification, prompt-driven rules, the suggested replies in the viewer — all run through a Large Language Model. The app license covers the features themselves, but the model has to come from somewhere. There are three ways to provide one, and you can mix them.
The three options at a glance
| Option | Cost | Privacy | Setup effort | Performance | Best for |
|---|---|---|---|---|---|
| Local LLM (Ollama, LM Studio, llama.cpp) | Free — your hardware | Highest. Nothing leaves your machine. | Highest. Install a server, pull a model, point Saymail at it. | Depends on your CPU/GPU and the model size. | Privacy-first users with a capable machine. |
| Bring your own key (OpenAI, Anthropic, Google, OpenRouter, OpenAI-compatible) | Pay-per-use directly to the provider | Whatever the provider’s policy is. | Medium. Get an API key, paste it in, pick a model. | Best. Top-tier hosted models. | Power users who already have a provider account or want fine-grained model control. |
| Saymail Cloud credits | Pay-per-use to Saymail | Routed through Zero-Data-Retention providers; nothing is kept beyond the request. | Lowest. Buy credits on saymail.eu — Saymail handles the rest. | Same quality as bring-your-own-key. | Users who want one bill, no API-key management, and don’t run their own hardware. |
How to switch between options
Open the LLM Connections panel from the command palette or panel menu. Each connection — whether it’s a local server, a provider key, or the managed Saymail Cloud entry — has a numeric Priority. Saymail uses the lowest-priority working connection first and falls back to the next if it fails.
You can have all three running at once: a fast local model as priority 1, a cloud provider as priority 2, and Saymail Cloud as priority 3. If your laptop is offline, the cloud takes over; if your provider key runs out of credit, Saymail Cloud picks up.
Read more
- Bring your own LLM key — how to add OpenAI, Anthropic, Google, OpenRouter, or any OpenAI-compatible service.
- Running a local LLM — Ollama and other local servers.
- Saymail Cloud — what credits are, where to buy them.
- LLM connection health — what to do when a connection fails.