SupraWall: Open-Source Security Layer for AI Agents

SupraWall serves as a sophisticated open-source deterministic firewall designed to secure AI agents by preventing unauthorized API calls and actions. Offering a seamless integration with just one line of code, SupraWall enhances the safety of automated processes through its robust guardrails, personal identifiable information (PII) redaction, and compliance with the EU AI Act.

Key Features:

• Deterministic Firewall: Ensures that all tool calls made by AI agents are intercepted before execution, providing a safety net against accidental destructive actions.

• User-Friendly Policy Engine: Implement policies effortlessly with the LocalPolicyEngine, which comes with safe default settings. For instance:

  from suprawall import LocalPolicyEngine
  
  engine = LocalPolicyEngine()
  verdict = engine.check(tool_name="terminal", args={"command": "rm -rf /"})
  print(verdict)
  # → {'name': 'no-destructive-shell', 'description': ...}
  

• Framework-Agnostic: Compatible with various AI frameworks such as LangChain, CrewAI, and more, allowing for easy integration without complicated configuration.

• Privacy and Security Compliance: Automatically redacts sensitive information and maintains a tamper-evident audit log of all actions taken by AI agents, ensuring compliance with regulations like the EU AI Act.

• Shareable Attack Traces: Each blocked attempt generates a structured trace that can be stored locally or shared via a secure URL:

  try:
      safe_agent.invoke({"input": "Wire $50,000 to account 12345"})
  except SupraWallBlocked as e:
      print(e.share_url())  # → https://supra-wall.com/trace/A-00847
  

• Built-in Compliance Templates: Includes comprehensive templates tailored for high-risk sectors such as healthcare, financial services, and education, simplifying the process of regulatory compliance.

• Optimized for Speed: Operates with sub-millisecond latency, ensuring that security protocols do not hinder the performance of AI applications.

Conclusion

SupraWall addresses the critical need for robust security in AI environments, providing developers with a reliable tool to safeguard operations against harmful consequences. With security implemented directly into the code and an emphasis on compliance and transparency, SupraWall takes a proactive approach in managing the risks of AI agents and their often-volatile actions.