# Praxis AI

A set of components for building an AI gateway on top of the Praxis proxy, adding provider-aware routing, protocol translation, response storage, guardrails, and observability at the proxy layer.

Praxis AI extends the Praxis proxy framework with components for building an AI Gateway: it routes AI traffic by model, provider API format, or tool composition; translates between provider protocols, including OpenAI Responses and Conversations and Anthropic Messages, with streaming event translation; and supports agentic workloads by classifying and brokering MCP and A2A traffic. Gateway features include credential injection, prompt enrichment, external guardrails, token usage accounting, and response storage in SQLite or PostgreSQL.

It is kept in a separate repository from the core praxis-proxy/praxis framework so that the core proxy has no dependency on AI-specific code and can keep supporting standard, non-AI use cases; the project says this also lets AI capabilities evolve faster than the more mature core networking proxy. It is extensible through custom Rust filters built on Praxis's HttpFilter interface.

Website: https://github.com/praxis-proxy/ai