I built TrustLayer because I was spending $200+/month across Claude, GPT-4, and Gemini with zero visibility into what I was actually paying for — and no way to know which AI gave the right answer when they disagreed. The moment that pushed me over the edge: I published a stat from an AI that turned out to be completely hallucinated. Confidently wrong, zero warning signs. So I built a trust layer that sits between you and every AI you use. It scores every output 0-100 for hallucination risk, tracks costs across all providers in one dashboard, and now in v0.2 it works as an OpenAI-compatible proxy — set one env var and all your tools (Claude Code, Cursor, Aider) automatically flow through it. The biggest surprise during development: smart routing alone saves ~90% on simple queries by sending them to cheaper models automatically. Most "what is X" questions don't need GPT-4o. 100% local, MIT license, pip install and go. Would love feedback on what would make you use this daily.

TrustLevel AI

#AiSafety

#AiSafety