Hey Product Hunt! 👋 Ever suspected your AI gateway/relay was giving you a "watered-down" version of GPT-5.6? You pay premium prices, but the output feels... off. Maybe your system prompt gets ignored, or the token count doesn't add up. That's Model Downgrading (or "Token Dilution"), and it's rampant in the relay market. XTokenChecker is my answer to that. It’s a client-side integrity checker. Here’s how it works: You enter your API Key (stays in your browser). We send a crafted request to the relay, targeting a specific model (e.g., GPT-5.6). We analyze the response signature, headers, and token usage directly in your tab. 🛡️ What we catch: Model Swapping: Are they secretly routing you to deepseek-v3? Token Dilution: Are they adding hidden padding to inflate your bill? Protocol Tampering: Are they stripping out your system messages? Since everything runs locally, your key never touches our server. It’s a zero-trust way to keep your relay honest. Try the "Integrity Check" on your favorite gateway/relay and see if they pass the test! 🕵️‍♂️

XTokenChecker

Verify your AI gateway. Prove you’re getting the real model.

Verify your AI gateway. Prove you’re getting the real model.