SOUNDTRACK // IRON UNDER PRESSURE
JEZWEB // GRID OSv2026
JEZWEB

> mcp // email & communications

Spam Filter
for AI.

Centralised form submission spam filtering with AI-powered content analysis, email reputation checks, and business context matching.

Every contact form on a client website is a target for spam bots and low-quality lead submissions. This server runs a four-layer scoring pipeline (content analysis, email reputation, business context matching, and Workers AI for borderline cases) before a submission ever reaches the CRM. Your AI assistant can check submissions, audit email addresses, review filtering stats, and feed back on false positives.

GOOD FOR // Web agencies managing contact forms across multiple client sites who want AI-assisted spam filtering that improves over time without tuning each site separately.

Form submission screening

Pass any inbound form submission through the full scoring pipeline. Gets a spam verdict with a confidence score before it is forwarded or saved anywhere.

Standalone email reputation checks

Check an email address against reputation signals without processing a full submission. Useful for validating a contact before adding them to a CRM or list.

Feedback and continuous improvement

Mark a decision as spam or legitimate to improve future scoring. Feedback is idempotent so running it twice on the same submission is safe.

Filtering statistics

Pull per-domain filtering stats for any time period: see how many submissions were blocked, how many passed, and where the borderline cases landed.

Check this form submission from info@quickloans-promo.com. Is it spam?

What is the spam rate for clientsite.com.au over the last 30 days?

Mark submission ID 4821 as a legitimate enquiry. It was from a real customer.

Check the reputation of this email address before I add it to the newsletter list.

check_submission Primary spam check for form submissions readOnly: false, openWorld: true
check_email Standalone email reputation check readOnly: true, openWorld: true
report_spam Record feedback (spam/legitimate) readOnly: false, idempotent: true
get_stats Filtering statistics by domain/period readOnly: true, openWorld: false

Works with Claude (Code, Desktop, claude.ai), and any MCP-compatible client. OAuth on first connect: no keys to paste.

claude mcp add --transport http spam-filter-mcp https://spam.mcpserver.au/mcp

claude.ai: Settings → Connectors → Add connector → https://spam.mcpserver.au/mcp