Humane Proxy

Humane Proxy is an MCP (Model Context Protocol) server providing high-trust AI safety middleware for any LLM-backed application. Its main function is to detect and intercept prompts indicating self-harm or criminal intent before they reach an LLM agent, offering empathetic care responses and alerting operators as needed. It can be used as a plug-and-play middleware proxy, Python library, or as a standalone MCP server, with configurable 3-stage safety pipelines powered by heuristics, semantic embeddings, and LLM-based reasoning. Integrations include OpenAI, Groq (LlamaGuard), and any upstream LLM via API, making it ideal for teams, developers, and organizations building safe, responsible AI assistants and platforms.