Hey Product Hunt, Most "private AI" claims are a paragraph on a privacy page. We wanted something you could actually verify, so we built SayGM's gateway to run inside an Intel TDX enclave; the chip itself signs a measurement of exactly what code is running. It either matches our open-source release or attestation fails. That's the proof, not our word. What that gets you: one OpenAI-compatible endpoint for Claude, GPT, and Gemini. Drop it into Cursor, Cline, Claude Code, or the Vercel AI SDK with no rewrite. For closed models, your prompt is sealed in the enclave so only the model maker you picked sees it, never us or the host. For open models, we run inference inside the enclave too - not even the model maker sees the prompt. It's built on Bittensor, and logging in gets you a discount, cheaper than paying the model maker directly or routing through OpenRouter. We've recently launched and would love feedback - especially from anyone who's been burned by silent price hikes, surprise model cutoffs, or "trust us" privacy claims from other gateways. Say gm 👋

SayGM

Say gm to AI | One private API for Claude, GPT & Gemini

Say gm to AI | One private API for Claude, GPT & Gemini