Skip to content

Lower your clients’ AI costs. Get paid when you do.

Built for agencies and freelance AI developers who need reliable, affordable model access — and want to earn from recommending it.

Get API Key — $5 free creditSee the referral program

A key per client

Separate keys, separate spend caps, separate logs. Bill each client for exactly what they used, with an itemised invoice to prove it.

Margin you control

Keep the difference as margin, or pass it through as a competitive advantage in your own pricing. Your call, per engagement.

Licensing you can defend

Every model we serve is open-weight under its own license. Nothing in your client's stack depends on a gray-area resale arrangement.

How the numbers usually work

Example scenario

An AI automation studio builds a customer-service assistant for an e-commerce client, running roughly 400M tokens a month on GLM-5.2. At official pricing that project costs about $952/month. Through AI Token Router the same workload costs about $551/month — $401 of monthly margin they can either keep or pass along.

This is an illustrative example, not a verified customer result. The token volume and split are assumptions; the rates are the real ones from our pricing page, so you can check the arithmetic yourself.

Earn 10–15% recurring commission

On every client you bring to AI Token Router, for 180 days per referral.

See how it works

Running several clients on one account

The operational mechanics, since an agency's problem is rarely the rate and usually the bookkeeping.

One key per client engagement

Keys are created and revoked from the dashboard, and a key is shown in full exactly once at creation. Per-client keys are what make usage attributable at invoicing time and what let an engagement end cleanly: revoke the key, and that client's access stops without touching anyone else's.

The balance is the ceiling, for everyone at once

Spend is prepaid and enforced on every request, so the account balance bounds every client project on it simultaneously. A client whose agent loops overnight stops at what you have already paid rather than producing an invoice. Automatic recharge is off unless you turn it on, and when on it is capped at three recharges in any 24 hours and disabled after three consecutive declines.

Rebill or pass through, per engagement

Our rate is published per model with the official rate beside it, so both models of agency pricing survive contact with a client who checks: keep the difference as margin, or pass it through and show the saving as part of what you delivered.

The two relationships are separate

Using the API for client work and referring clients to it are different accounts and different mechanics. A client you refer signs up in their own name, pays their own bill and keeps their own keys; you earn 10–15% of what they spend for 180 days without sitting in the middle of their billing.

Scope a client project against what is callable today

Seven of twenty-four catalogued models are served now. That is the number to design against.

Kimi K2.6, Kimi K3, GLM-5.2, DeepSeek V4 Pro, DeepSeek V4 Flash, Qwen-Image, Qwen3 Embedding 8B are callable today. The remaining models — including every video model in the catalogue — are published with their rates and marked unavailable until an upstream serves them, so a request for one returns no available channel rather than quietly routing somewhere else. Two more constraints worth putting in a proposal: there is no closed frontier model here, by licensing choice rather than roadmap, and there is no automatic failover between models, so redundancy is something you design rather than something you inherit.

Availability is marked on every model page and returned by GET /v1/models.

Questions agencies ask

Can I keep each client's usage and billing separate?

Yes — create one API key per client engagement. Keys are created and revoked in the dashboard, which is what makes usage attributable per client and what lets an engagement end cleanly. Spend across all of them draws on one prepaid balance, and that balance is a hard ceiling enforced on every request, so no client project can produce a surprise invoice.

Should I resell at our rate or pass the saving to the client?

Either works, because our rates are public per model with the official rate printed alongside. Keeping the difference is margin on infrastructure you are already managing; passing it through is a competitive argument in your own pricing. The one option that does not survive is quoting a rate that differs from the pricing page, since the client can read it.

What can I safely promise a client about model availability?

Promise what is callable now: seven of the twenty-four catalogued models are served today, including GLM-5.2, Kimi K3, Kimi K2.6 and both DeepSeek V4 tiers. The rest, including all video models, are listed with published rates and marked unavailable until an upstream carries them. There is also no automatic failover between models, so redundancy is a design decision in the client's architecture rather than something the gateway does for you.

Do I earn on referrals and use the API myself at the same time?

Yes, and they are separate relationships. A referred client signs up in their own name and pays their own bill, while you earn 10–15% of what they spend for 180 days from the referral, with a 60-day window between their first click and their signup. The commission is paid by us and never added to their bill.

What licensing exposure am I handing my client?

Only open-weight models are served here, each under its own licence, which permits third-party hosting — so nothing in the client's stack depends on a gray-area resale of closed-weight access. Each model's licence is printed on its catalogue page and returned by GET /v1/models, which is usually enough for a client's legal review to close the question.

Talk to us about client billing

Per-client attribution, consolidated invoicing, or anything the pages above did not answer.

We respond within 24 hours.

Start with one client project

$5 in free credits, no credit card required.

Get API Key