SayGM is a cheaper, private alternative to OpenRouter Most private AI claims are a paragraph on a privacy page. SayGM proves it instead: the gateway runs inside an Intel TDX enclave, so the chip itself signs a measurement of exactly what's running. One OpenAI-compatible endpoint gets you Claude, GPT, and Gemini, no rewrites. Closed models stay sealed for the maker alone; open models run inside the enclave too, so no one sees your prompt.