We are MailTag.
Your inbox, sorted every 5 minutes.
Privately.

Paste an email. Watch it pick.

This is the real classifier running on our servers, with the default categories. Nothing you type is stored.

It doesn't write. It picks.

Most "AI inbox" tools send your email to a chatbot and hope the reply contains a valid label. MailTag uses the OpenJev "semantic if" technique: the model reads the probability of every category you allow in a single forward pass, and the highest one wins.

• Can't hallucinate a label. The output is always one of your categories — there is nothing else it can return.
• Deterministic. Same email, same category. No temperature, no prompt roulette.
• A real confidence score. Every label comes with a probability, not a guess dressed up as certainty.
• Fast and cheap. One forward pass instead of generating text — that's how we can price at $9.
• Open-weight models we host. Your email is never sent to OpenAI, Google or any third-party AI API.

Route the important stuff where you work.

Attach an outgoing webhook to any category. Send every Urgent email to Slack, every Receipt to your bookkeeping Zap, every Needs reply to your task list.

• Per-category webhooks on every plan — works with Slack, Discord, Zapier, Make and n8n.
• Signed requests. Every delivery carries an HMAC-SHA256 signature in the X-MailTag-Signature header.
• Triage API (Pro). Classify any email over HTTP — plug it into Postmark inbound, Cloudflare Email Workers or Mailgun routes.

Request

curl -X POST https://mailtag.app/api/v1/triage \
  -H "Authorization: Bearer $MAILTAG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "from": "anna@acme.com",
    "subject": "Prod is down for EU customers",
    "text": "Can you jump on a call right now?"
  }'

Response

{
  "category": "Urgent",
  "confidence": 0.93,
  "scores": {
    "Urgent": 0.93,
    "Needs reply": 0.05,
    "FYI": 0.01,
    …
  }
}