Secure your AI agent in 5 minutes
Sign up, install the SDK, and wrap your AI in three lines. No retraining, no proxy, no refactoring. Stop prompt injections and malicious tool calls with a single API call.
1# Install2pip install sonnylabs3
4# Protect in 3 lines5from sonnylabs import SonnyLabsClient6
7client = SonnyLabsClient(8 api_token="your_token",9 analysis_id="your_id"10)11
12# That's it - you're protected on the input13result = client.analyze_text(14 user_input,15 scan_type="input"16)Integration in 3 Simple Steps
From zero to production-grade AI security in under 5 minutes. No complex configuration, no code refactoring.
Install
⏱️ < 1 minute
Add SonnyLabs to your project with a single command
pip install sonnylabsNo credit card required • Free tier available
Before & After SonnyLabs
One function call stands between your AI and catastrophic security breaches
WITHOUT SonnyLabs
Vulnerable to attacks
# Without SonnyLabs - Vulnerable!
response = ai_model(
user_input # ❌ No protection
)
# Any malicious input goes straight to your AI
# Prompt injections, jailbreaks, data exfiltration...Security Risks:
WITH SonnyLabs
Production-ready security
# With SonnyLabs - Protected!
result = client.analyze_text(
user_input,
scan_type="input"
)
if result['clean']:
response = ai_model(
user_input # ✅ Verified safe
)
else:
# Handle malicious input safely
response = "Request blocked"
# Real-time protection in one function callSecurity Benefits:
Works with any AI architecture
The same SDK protects agents, chatbots and MCP servers. Sit in front of OpenAI, Anthropic, Gemini or your own model.
AI Agents
Wrap inputs and tool calls for agents that take real actions.
Chatbots
Protect customer-facing conversational AI on the way in and out.
MCP Servers
Guard Model Context Protocol tools against injection and abuse.
Real-Time AI Firewall for Your Applications
Born from cutting-edge research at one of Ireland's top universities, SonnyLabs is a proprietary AI security solution that detects and blocks threats like prompt injections in real-time, protecting your AI applications before malicious inputs reach your models.
EU AI Act compliant. Excellent data privacy since it's not based on OpenAI or Anthropic.
Flexible deployment: Use our API or self-host for complete control.
Choose Your Model
Proprietary AI security models to detect and block threats like prompt injections in real-time, optimized for speed or accuracy based on your needs.
Speed Optimized
Lightning-fast threat detection for real-time applications requiring instant responses.
Accuracy Optimized
Maximum precision threat detection for critical applications where accuracy is paramount.