Skip to main content

​
What is Velobase?

Velobase is an AI gateway that meters and bills every model call to your end-customers. Point your OpenAI- or Anthropic-style SDK at https://api.velobase.io/v1, add one header for who to bill (X-Velobase-Customer: <user-id>), and every model call is metered against that customer’s wallet. The cost comes back on the response headers (x-velobase-cost-cents, x-velobase-cost-credits, x-velobase-balance-credits). It is the per-customer wallet, ledger, and rate-limit layer OpenRouter leaves you to build. It is also a high-performance billing ledger, so you can meter work you price yourself against the same wallets.
Using an AI coding agent? Install the MCP server to let your agent write the integration for you, no manual setup needed.

​
How it works

1

Get an API key

Sign in at velobase.io/dashboard and generate a key in your project’s Keys section. A project key (vb_live_) runs server-side and names the customer to bill via the X-Velobase-Customer header; a customer-scoped key (vb_customer_) carries its own customer binding.
2

Fund a customer

Call POST /v1/customers/deposit with a customer_id and an integer credit amount. Velobase creates the customer automatically if they don’t exist yet.
3

Make a billed model call

Point your OpenAI or Anthropic SDK at https://api.velobase.io/v1, call POST /v1/chat/completions or POST /v1/messages, and send the X-Velobase-Customer header. The call is billed to that customer’s wallet and the cost comes back on the response headers (x-velobase-cost-cents, x-velobase-cost-credits). For work you price yourself, use direct deduction or the freeze then consume flow.
4

Track every transaction

Every operation is recorded in an immutable ledger. Read a customer’s wallets with GET /v1/customers/{id} and reconcile usage in real time with GET /v1/usage.

Get started in 5 minutes

Make your first model call, fund a customer wallet, and bill a request end to end.

​
Key features

Model gateway

Point your OpenAI or Anthropic SDK at Velobase and bill every model call to a customer wallet. Live models include deepseek/deepseek-v4-pro (streaming) and the anthropic/claude-* family (Bedrock, non-streaming), with more available via upstreams.

Customer deposits

Wallet categories, validity windows, and auto-provisioning for customer wallets.

Direct deduction

Charge immediately when the cost is known upfront: image generation, fixed-price APIs.

Staged deduction

Freeze a budget, then settle the actual cost: LLM inference, async tasks.

Immutable ledger

Every wallet change is recorded permanently with a full audit trail.

​
Billing flows for work you price yourself

Beyond the model gateway, the same wallets back a transaction-based ledger for anything you price yourself. Customer deposits use integer credits via amount; other billing primitives may use explicit unit fields such as amount_usd, amount_cents, or amount_credits. 1 USD = 100 cents = 1,000,000 credits.

Direct deduction

Fixed cost known upfront via POST /v1/billing/deduct

Freeze then consume

Variable cost settled after via POST /v1/billing/freeze then /v1/billing/consume

Deposits

Top up customer wallets via POST /v1/customers/deposit

​
Why Velobase?

Stripe is built for subscription and invoice workflows. AI workloads need sub-second credit operations with atomic freeze-and-settle semantics. Velobase handles thousands of concurrent operations per second with zero race conditions and no negative balances.
When you freeze funds, Velobase atomically reserves them from the wallet’s available balance. No other request can spend those funds while they’re frozen. After the task completes, settle the actual cost and any unused amount is returned automatically.
Every billing operation is idempotent. Retry with the same transaction_id and Velobase returns the original result without double-charging.
Yes. The direct deduction flow works for any fixed-price operation: game currencies, API rate limits, marketplace points, or anything you bill against a customer wallet.

​
Explore

Quickstart

First API call in under 5 minutes

API Reference

Complete endpoint reference

MCP Server

AI coding agent integration

Dashboard

Overview, Models, Customers, Keys, Usage, Settings

Idempotency

Prevent double charges

Error Handling

Error codes reference

Website

velobase.io

Dashboard

Console

GitHub

Source & issues