Show HN: Open Bias – proxy that enforces agent behavior at runtime
Category: security
Tags: llm-agents, proxy, guardrails, ai-safety, runtime-enforcement, policy-engine
Score: 7.3/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 7)
Open Bias is an open-source proxy that sits between applications and LLM providers to enforce behavioral rules defined in a simple RULES.md file at runtime. It offers multiple enforcement actions like blocking, intervening, or shadowing, and supports pluggable evaluation engines, making agent alignment practical and configurable without heavy prompt engineering.
Target audience: backend devs, ai engineers, devops
Repository: https://github.com/open-bias/open-bias/ · Python · Apache-2.0 · 71 stars
View on Hacker News