Torrix

Most LLM observability tools send your prompts to their cloud. Torrix runs on your server. Add 2 lines of Python, or route any HTTP client through the proxy. No code changes needed. Every AI call is logged instantly: tokens, cost, latency, and the full prompt trace. Works with OpenAI, Anthropic, Gemini, Groq, Azure, Mistral, SAP AI Core, n8n, and any HTTP API. Community edition is free forever. Your data never leaves your infrastructure.