Liquid neural networks detect malicious traffic patterns in real time. No signatures to maintain, no cloud calls, no privacy compromise. One Docker command, then your APIs are protected.
Ward learns the natural patterns of your real users (timing, navigation, interaction rhythms) and flags everything that deviates. Zero-day bots are caught the moment they hit the API.
Drop Ward in front of any API. After ~500 requests it has learned your traffic baseline and switches to detection mode autonomously. No labels, no manual rules, no DevOps overhead.
Ward runs on the same server as your API and returns a verdict in microseconds. Zero round trips, zero cloud dependencies, zero added latency for your real users.
Every request stays inside your infrastructure. No third-party processing, no data transfer agreements, no cookie banners required. EU compliance by design.
Powered by our Selective State Space engine — the same technology used to detect anomalies in industrial sensors at microsecond scale. Battle-tested under real production load.
Standalone Flask service on port 9741. Call it from any backend (Node, Go, PHP, Python, Rust) with a single HTTP POST. No SDK to integrate, no framework lock-in.
Start the Docker container on your server or cluster. Ward listens on port 9741.
Your API forwards request metadata (IP, UA, method, path) to Ward via HTTP POST.
Ward observes the first ~500 requests to build a baseline of legitimate traffic patterns.
From request 501 onwards, every call gets a verdict in 0.2ms. Block, challenge, or log.
Self-hosted, no signup needed for the open-source build. Production licenses available.
Self-hosted bot detection that respects your users' privacy and your servers' uptime.