Outbound only
Each app opens an outbound WireGuard tunnel to the mesh. No port-forward, no static IP, no inbound firewall rule.
Put NYXR in front of any app or API and it stops attacks, bad bots and abuse before they ever reach you. Nine coordinated layers of defense, an AI that makes the final call, and full control you can explain and undo - all self-hosted.
Every request flows through the full pipeline in order. A single bypassed control is never enough, and the final AI verdict is authoritative all the way down to the kernel.
Tap any layer to see exactly what it does.
NYXR continuously turns real traffic into service-specific baselines, fingerprint reputation and emerging campaign detections. The learner adapts without becoming a black box: high-impact rules wait for your approval, remain explainable and can be rolled back.
Correlate JA4, header order, client coherence and behaviour across rotating IPs to recognize the same automation stack.
Cluster suspicious fingerprints, autonomous systems and attack ratios before a low-and-slow botnet becomes an incident.
Learning runs continuously, but block and ban proposals stay reviewable. Approve, reject or roll back every learned rule.
The adaptive bot-detection engine: learned traffic baselines, client fingerprint reputation, discovered campaigns and the learned-rule approval queue (shadow rules auto-promote for soft actions; block/ban land as proposals to approve).
Stable client stack observed across 38 IPs with a 91% attack ratio and incoherent browser signals.
17 IPs from one ASN are targeting /wp-login.php across 6 services.
Signatures and scores resolve most traffic. For the ambiguous remainder, an AI verdict engine weighs the full request context and returns a single, accountable decision the rest of the stack obeys.
Every verdict carries the signals and the reasoning behind it - never a silent black box.
An AI "allow" lifts a behavioural auto-ban, and a block reaches down to the kernel nftables layer.
Any decision can be overridden and rolled back from the console in one click.
Real-time security posture, traffic and alerts across every protected service.
Active bans and the AI verdicts that drive them. Enforcement and intelligence in one view.
Search and stream WAF security events - filter by action, rule and client IP.
Publish apps running at home or in a private network without a static IP, a port-forward or a hole in your firewall. NetBird joins them to NYXR over an encrypted WireGuard mesh, and only NYXR faces the internet.
Your private network
Your apps stay behind NAT. Each one dials OUT to the mesh, so nothing inbound is ever opened.
Each app opens an outbound WireGuard tunnel to the mesh. No port-forward, no static IP, no inbound firewall rule.
NetBird links peers over an end-to-end encrypted overlay, so your origin stays private and unreachable directly.
NYXR is the sole internet-facing entry: it filters, runs the WAF, and reaches your app across the mesh.
The data plane reads compiled snapshots and never runs a per-request database query. The control plane versions, validates and atomically swaps every change, so a bad config can always be rolled back.
The Hono API versions every config change, validates it, swaps it atomically and can roll it back. No edit ever ships unchecked.
The OpenResty + ModSecurity gateway reads compiled snapshots locally and never runs a per-request database query. Fast and fail-safe.
Dedicated workers refresh threat feeds, process events, run backups and dispatch notifications, fully decoupled from the request path.
ClickHouse, Prometheus and Grafana feed a real-time, fully bilingual Astro console where every security decision is searchable and explainable, alongside live health for every container in the stack.
Self-hostable, fully observable, and reversible by design. Bring NYXR in front of any service and neutralize the risk at the edge.