Your first detection is one pip install away
Free during early access. The full multi-signal detection engine, running locally in your process. No API key required – prompts stay in your boundary.
Three steps to runtime detection
Install
pip install raxeIntegrate
from raxe import Scanner
scanner = Scanner()
result = scanner.scan(prompt) # verdict before the model call
if not result.threat_detected: # result.action: ALLOW / LOG / REVIEW ...
response = llm(prompt)Observe
[RAXE] Threat detected: prompt_injection
severity: critical
technique: AML.T0051 (Direct Prompt Injection)
action: logged with evidence (observe-first – the would-block decision is recorded)Illustrative output – labelled sample
Verdicts are explainable: threat probability, per-family scores, out-of-distribution signal, and the nearest known attack patterns – not just a label.
The agents on your laptop are a surface too
The same stack watches coding agents as they work – no SDK call required, and attribution is labelled honestly in the console.
Test detection quality in your browser
A browser-only demo of the text classifier. Throw real attack patterns at it – nothing to install, nothing leaves your machine.
Open the Detection LabWorks with your framework – no adapter required
Every Python agent framework hands you the text somewhere – a callback, a hook, a tool wrapper. That is the integration point. No purpose-built adapter required.
# wherever your framework exposes the prompt or tool input –
# a LangChain callback, a LiteLLM hook, a CrewAI tool wrapper –
result = scanner.scan(tool_input) # action, threat_detected, confidence, familyIf the generic pattern does not fit your stack, tell us – design partners set the adapter queue.
What you get
5 Detection Signals
ATLAS + ASI-mapped rules, ML text scorer, tool-call risk analysis, agent behaviour graph, optional advisory LLM judge
Open Signatures
Open, auditable detection signatures – every one mapped to MITRE ATLAS and OWASP ASI
On-Device ML
ONNX inference, no GPU required, runs locally in your process
Local Latency
Fast rules answer in single-digit milliseconds; the full multi-signal verdict in ~150 ms – all local
13 Threat Families
41 techniques, 10 harm types – one taxonomy across every signal
Observe-First
Deploys observe-and-log. Every would-block decision is logged with evidence – nothing in your call path changes behaviour
pip install today is RAXE Sensor [sdk] – the
in-process library form of RAXE Sensor, and the same detection engine that powers the
full self-hosted stack. Start with one scan call in one file; grow into a gateway, a
console, a kernel-level host sensor, and AWS cloud detection without changing
detectors.
pip install – today
Base-URL change – today
Kernel eBPF – today
DaemonSet / sidecar – roadmap
Docker Compose + systemd + Python SDK are the validated deployment shapes.
Early access – design partner programme. Free during early access. Ready to see it on your whole stack?