KireiFilter is a server-side API that scores incoming content — comments, contact form submissions, registrations, email — and tells you whether it's spam. No CAPTCHA. No friction for real users.
spam score
How it works
Send the content, IP address, and optional metadata to POST /api/v1/spam-check. KireiFilter runs it through multiple detection layers and returns a score between 0.0 and 1.0, a verdict, and which signals fired.
You decide what to do with the score — auto-reject, queue for review, or let it through. The logic stays in your code; we just provide the signal.
{
"verdict": "clean",
"score": 0.12,
"signals": [
{ "layer": "heuristic", "score": 0.12 }
],
"checked_at": "2025-03-18T09:14:22Z"
}Features
Known bad IPs and domains are hard-blocked immediately. Configurable per account — add your own blocklist entries.
Pattern-based scoring detects common spam signals: keyword density, link counts, excessive caps, suspicious phrases, and more.
Real-time blocklist lookups against public RBL providers to identify IPs with a history of sending spam or abusive traffic.
Measures character-level randomness to catch machine-generated gibberish, keyword stuffing, and auto-spun content.
Identifies repeated spam campaigns by hashing and comparing content. The same template sent by different bots gets caught the second time.
Detects disposable addresses, normalized duplicates, and email-as-password patterns that signal throwaway accounts and fake registrations.
Use cases
Score every submission before it hits your database. Auto-approve clean content, hold grey-area posts, reject obvious spam.
Block bot floods without CAPTCHA. Legitimate users see nothing different — spam never reaches your inbox.
Catch fake signups, disposable emails, and bot accounts at the point of registration before they pollute your user base.
Classify raw email content, headers, and sender data before processing. Useful for support inboxes, ticketing systems, and forwarding pipelines.
Detection pipeline
Each request passes through independent detection layers. Scores are accumulated, high-confidence matches short-circuit the chain. You get the final score and every signal that fired.
Principles
Spam filtering happens on your server. Real users never solve a puzzle, click a checkbox, or prove they're human.
Data stays in Europe. No third-party trackers, no cross-account data sharing, no surprises at audit time.
We return a score and signals. What you do — reject, queue, approve — stays in your code. No magic auto-blocking outside your control.