Three things that change everything.
One key, any model
Claude Haiku, Sonnet and Opus alongside the GPT-5.6 family — behind a single credential and one request shape.
How it worksOne prepaid wallet
No subscription, no seats, no balance stranded across four provider accounts. One pot funds everything.
See pricingBuilt for UPI
Top up on local rails and we settle with providers in USD. No international card required.
Why TokenFlowThree steps to billing by the token.
Top up your wallet
UPI or card, from ₹50. Prepaid, never expires.
Chat or build with one key
The browser app or your own code — same wallet either way.
Pay per token
Sub-cent precision, every charge logged.
Frequently asked questions
Is there a subscription?
No. TokenFlow is pay-per-token — you top up a prepaid wallet and pay only for what you use, metered to sub-cent precision. No monthly plan, no seats, no minimum.
How does pay-per-token pricing work?
Every request is billed the underlying provider cost plus a flat, transparent markup. Both the raw cost and your charge are logged, so your spend is always auditable. You never pay for unused quota.
Which AI models can I use?
One TokenFlow key reaches the Claude and GPT families. Send model="auto" to let TokenFlow route each request to a model that fits (cheaper for simple prompts, stronger when needed), or pin a specific model yourself.
Can I pay with UPI?
Yes. TokenFlow is built for emerging markets — top up in ₹ via UPI or card (Razorpay), and we handle paying the AI provider in USD behind the scenes.
Do I need a provider account or API key?
No. You use a single TokenFlow key — no Anthropic or OpenAI account, no provider SDKs, no per-provider keys, and one consistent response shape whichever model answers.
Is there a free way to try it?
Yes — 20 free messages a day on GPT-OSS 20B, no card needed, shared across the chat app and the API and reset every day. It rises to 100 a day once you’ve topped up. Your first top-up — from ₹50 or $2 — is also matched with a small bonus credit.