Show HN: OpenAPPA – open-source deterministic guardrails that don't break agents

Category: security

Tags: ai-security, guardrails, agentic-ai, information-flow-control, rust

Score: 8.3/10 (Innovation: 8, Technical: 8, Documentation: 9, Utility: 8)

OpenAPPA is a deterministic guardrail system that sits between LLM agents and their tools, using an Agentic Permissions Policy Algebra to track data sensitivity and trust across tool calls, thereby preventing data exfiltration via prompt injection or hallucination. It offers a declarative TOML policy, a deterministic engine that derives decisions from event logs, and machine-readable remedy plans to preserve agent utility. Backed by a NeurIPS workshop paper and benchmarks showing zero successful attacks, it is an innovative and practical security layer for AI agents.

Target audience: AI/ML engineers, security engineers, and platform teams deploying LLM agents

Repository: https://www.openappa.com/ · Rust · MIT · 1598 stars

View on Hacker News