The most common question we hear is: "Does Premail read my email?" It's a fair concern. Email is personal. The answer is no — and this post explains exactly why not.

What Premail actually sees

Premail only ever looks at the envelope of a message — the sender address and subject line. It never opens the message body. This is by design, not a limitation.

When a new message arrives, Premail reads two fields: From: and Subject:. That's it. These are the same fields you see in your inbox before you open an email. Nothing more is accessed.

How the AI classifier works

Premail sends the sender and subject to a local or cloud AI classifier. The classifier returns a category — things like "recruiter outreach," "newsletter," "transactional," or "genuine message" — along with a confidence score.

If the confidence is high enough and you have a rule for that category, Premail takes the configured action: move to a folder, archive, trash, or quarantine. The message body is never touched.

Local-first by default

Premail is built as a macOS app, not a cloud service. Your email credentials live in macOS Keychain. Your rules live on your machine. Classification happens locally when you use Ollama, or via a direct API call to Anthropic — a call that contains only the sender and subject.

There is no Premail server in the middle. We never proxy your email. We never store message metadata.

Why this matters

Most email filtering tools operate as a man-in-the-middle: your email routes through their servers, they scan the full content, and they decide what you see. This approach gives them enormous access to your private communication.

Premail takes the opposite approach. We filter at the metadata layer, which is sufficient for catching 95% of unwanted email, while keeping your message content entirely private.