Hi PH 👋 I built TokenFlow because paying for AI is harder than using it if you are not in the US or EU. The major APIs expect an international card and a foreign-currency charge, and that stops people before they write a line of code. One API key reaches seven models across Anthropic, OpenAI and Groq. You pay per token from a prepaid wallet, not a monthly plan. A quiet month costs nothing. Three things I would defend: 🔍 The markup is one public number. 15% over what the provider charges us, and both figures are written to your ledger on every call, so you can check it rather than trust it. 🧭 Auto-routing costs you nothing. Send model="auto" and a short classifier call picks a model that fits, so a simple question is not billed at flagship rates. That classifier call is one we absorb, logged at zero on your ledger. ⚡ A provider error before the first token is not billed at all, and fails over once to a different provider. Being upfront about the limit: top-ups are India-only right now, through Razorpay and UPI. Other regions are next. The 20 free messages a day need no card anywhere, so you can try it properly wherever you are. I would rather hear what is broken than what is nice.

TokenFlow

Pay-per-token AI. One key, seven models, UPI top-ups.

Pay-per-token AI. One key, seven models, UPI top-ups.