Waterfall Logo
Waterfall

Turn AI Usage

Into a Business

Model

Credit infrastructure for platforms building on LLMs. Give your users wallets, meter every model call, and settle per request — so you can ship usage-based AI products without building the billing stack.

OUSDS 0.001
AUSDS 0.005
DUSDS 0.021
GUSDS 0.013

Every LLM call settles against a stablecoin-backed wallet. Per-request. In real time.

Meter any LLM provider out of the box

OpenAI
Anthropic
DeepSeek
xAI

300+

Models Supported

Per-Request

Cost Attribution

Real-Time

Stablecoin Settlement

~150ms

Payment Overhead

The Problem

Building usage-based AI pricing is a full-time job

You want to charge users for what they actually consume. But metering, attribution, and settlement across LLM providers is harder than it looks.

Flat pricing bleeds margin

You're charging $20/mo but some users burn $200 in API calls. Usage-based pricing is the fix — but building the metering layer is a project in itself.

DIY metering is fragile

Logging costs after the fact, reconciling vendor invoices, hoping your numbers match. One missed call and your P&L is off.

No settlement primitive

There's no atomic link between "user made an LLM call" and "user paid for it." You're stitching together Stripe, webhooks, and prayer.

Three Steps

How It Works

Add usage-based billing to your AI product in minutes. Fund wallets, meter calls, settle automatically.

01

Fund User Wallets

Give each user, agent, or team a stablecoin-backed credit wallet. They fund it, you control the spend policy. No crypto complexity for your users.

02

Route LLM Calls

Use the SDK or gateway. Every request is metered by model, token count, and cost — attributed to the right user in real time.

03

Settle Per Request

Each API call settles against the user's wallet atomically. No batched invoices. No reconciliation. Credits go down, you get paid.

Our Edge

Not observability. Accountability.

Most tools track AI costs after the fact. Waterfall settles them at the moment they happen — using x402 and stablecoins so the blockchain is your accounting layer, not a dashboard you check on Mondays. We're not a crypto company — we use blockchain as infrastructure to make AI cost tracking exact, instant, and tamper-proof.

Traditional approach

Waterfall

Cost tracking

After the fact, batch analytics

At the moment of each request

Settlement

Monthly invoices + reconciliation

Atomic per-request via x402

Attribution

Org-level dashboards

Per-user, per-agent wallet ledger

Source of truth

Vendor invoices you hope are right

Immutable stablecoin transaction ledger

Billing integration

Stripe + webhooks + custom code

Built in — wallets are the billing

Overspend protection

Alerts after you've already overspent

Wallet balance enforced per call

Zero

Invoice reconciliation

Every LLM call settles atomically against a stablecoin wallet. There's nothing to reconcile — the ledger is the payment.

100%

Cost attribution accuracy

No sampling, no estimates. Every token is metered and settled at the request level, tied to a specific user or agent wallet.

Real-time

Revenue recognition

Credits deduct the moment a model responds. Your revenue is recognized instantly, not 30 days later on a vendor invoice.

Supported Providers

OpenAI
OpenAI
GPT-4, DALL-E, Whisper
Models

15+

Requests

4.2M

Anthropic
Anthropic
Claude 3.5, Claude 3
Models

8+

Requests

2.8M

DeepSeek
DeepSeek
DeepSeek V3, Coder
Models

6+

Requests

1.5M

xAI
xAI
Grok-2, Grok-2 Vision
Models

4+

Requests

890K

OpenAI
OpenAI
GPT-4o, o1-preview
Models

12+

Requests

3.1M

Anthropic
Anthropic
Claude Instant, Haiku
Models

5+

Requests

1.9M

DeepSeek
DeepSeek
DeepSeek Chat, Math
Models

7+

Requests

2.1M

xAI
xAI
Grok-1, Grok-1.5
Models

3+

Requests

650K

OpenAI
OpenAI
Embeddings, TTS
Models

10+

Requests

5.7M

Anthropic
Anthropic
Claude Opus, Sonnet
Models

6+

Requests

2.3M

DeepSeek
DeepSeek
DeepSeek R1, Lite
Models

4+

Requests

980K

xAI
xAI
Grok-3 Preview
Models

2+

Requests

420K