Enterprise

Every frontier model, one key, one invoice.

Metered access to 97 chat models through one OpenAI-compatible endpoint. API keys for your team, a ledger of exactly what each call used, and a bill in dollars, by wire or monthly invoice.

Nothing has to migrate first: the endpoint is OpenAI-compatible, and a Verified Eval on your own cases runs through it before you commit a workload.

What you get today

01

API keys

Access is on keys (brk_live_…) issued to your account. Point any OpenAI-compatible client at the endpoint and every chat model on the gateway is a model name away.

02

Exact usage ledger

Every call is metered on what it actually used and billed after the fact — no per-call minimum, nothing rounded up. The invoice is the ledger, totalled.

03

Monthly invoice, in dollars

Pay by wire — as prepaid credit drawn down at exact usage — or on a monthly invoice. Terms are agreed per contract.

How enterprises pay

Enterprise terms are agreed per contract rather than picked from a price list, and self-serve sign-in is not open yet: you talk to us, we issue keys, and metering starts with the first call. Usage is billed after the fact at exactly what each call used, with no per-call minimum.

Billing is in dollars, two ways: prepaid by wire and drawn down as you go, or a monthly invoice. Which one, and on what terms, is part of the contract.

The settlement model underneath the invoice — how each call is authorized, and what the gateway can and cannot claim — is documented in the security notes.

Security notes →

Talk to us

Tell us what you run and roughly what you spend, and you will hear back within two business days.

Support, billing and everything else: contact.

Enterprise AI gateway questions: keys, billing, models, migration

What do we need to start?

A conversation. Self-serve sign-in is not open yet, so we issue keys to your account directly and metering starts with the first call. No source, no data, no access to your systems is needed.

Which models can we call?

All 97 chat models on the gateway, through one OpenAI-compatible endpoint; switching models is changing a model name. See the catalogue →

How is usage billed?

After the fact, at exactly what each call used, with no per-call minimum. In dollars: prepaid by wire and drawn down as you go, or on a monthly invoice, as the contract sets out.

Do we have to migrate anything?

Not to start. The audit runs through our gateway on tasks you give us and hands back a routing and budget policy. Deploying it, and moving workloads onto keys, is a decision you make afterwards.

Does the enterprise gateway work with the OpenAI SDK, LangChain and our existing clients?

Yes. The endpoint is OpenAI-compatible and also speaks Anthropic's Messages shape, so an existing client points at it by changing the base URL and the key. Streaming, tool calls and vision behave as they do upstream.

Do you retain our prompts?

Yes, for abuse investigation and billing disputes. A prompt is sent only to the model that answers it and is not shared with anyone else. Each call's prompt and completion are kept in an access-restricted store, separate from the usage ledger.

Is there a per-seat price, a minimum commitment or a markup on tokens?

No seats and no per-call minimum: usage is metered at exactly what each call used, at the provider's list token rate. Commercial terms — prepaid by wire or a monthly invoice — are set per contract.

Enterprise BlockRun vs Portkey, LiteLLM or a provider's own enterprise plan?

A provider's plan covers one vendor; Portkey and LiteLLM route between providers whose accounts and keys you hold and are billed by each. Here one key reaches every vendor's models and one ledger bills them all in dollars — and the engagement starts with a measurement of what your agents actually spend, not a rate card. BlockRun vs Portkey →