Bring Your Own LLM: OpenAI, Ollama, or Any Endpoint

What it is

Bring Your Own LLM lets you choose what actually classifies your email: OpenAI, a local Ollama instance, or any OpenAI-compatible endpoint. AI Email Filter connects to whichever one you configure and sends it the classification requests.

The problem it solves

Most email AI tools decide the model for you, usually a cloud API you have no visibility into. That’s a fixed trade-off between cost, quality, and privacy that you can’t adjust. If you run your own mail server, or you simply don’t want email content going to a third-party API, a hardcoded LLM provider is a dead end.

How it works in AI Email Filter

• Point AI Email Filter at OpenAI, a local Ollama instance, or any endpoint that speaks the OpenAI-compatible API.
• With a local Ollama instance, no email content leaves your own infrastructure — classification happens on hardware you control.
• You pick the price/quality trade-off yourself: a stronger hosted model for accuracy, or a local model for cost and control.
• The rest of the product — Sender Smart Categories, Multi-Action Rules, History and Undo — works the same regardless of which LLM is behind it.
• AI Email Filter connects over IMAP/SMTP to Gmail and Google Workspace, Yahoo, ProtonMail via Bridge, Fastmail, Mailcow, or a self-hosted server (Outlook/Microsoft 365 support is coming soon via Microsoft sign-in) — the LLM you choose is a separate decision from which mailbox you connect.

Accuracy still comes down to how clearly your category prompts describe what belongs in each folder, and that holds regardless of which LLM is doing the classifying. Any sender can be re-categorized at any time if a particular model gets something wrong.

BYO-LLM is what makes AI Email Filter viable on a fully self-hosted setup — see self-hosted email sorting for how it fits with a self-run mail server like Mailcow. The free tier’s 200 classifications work with any provider you configure; see pricing for unlimited use.