Skip to content

One gateway for all your AI models, with spend per project

Why REM CloudHub puts every AI call behind one gateway: model names with automatic fallback, keys that are attached instead of copied, and budgets you can see per key, per project and per organisation.

3 min read

Most teams start using AI the same way: someone creates a provider key, pastes it into an environment file, and a month later nobody knows which project spent what, or which copies of the key exist. REM CloudHub's AI gateway is built to make that impossible by design. Here is how it works.

Call a name, not a provider

Your projects call REM CloudHub model names such as rch-fast and rch-advanced. Behind each name are one or more provider models in a fixed order. The gateway sends a call to the healthy route with the lowest order; if it fails or is cooling down, the next one takes over. Routes with the same order share the load. Your code never changes when a provider has an outage or when the platform moves a name to a better model, and your plan decides which names your organisation can call.

The gateway speaks the OpenAI chat-completions format, so any OpenAI client works: point it at the endpoint on your key's Use it tab and pass the key.

Keys are attached, never copied

An AI key in REM CloudHub has a name, the models it may use, a budget in dollars with a reset period, and limits on requests and tokens per minute. Its value is shown once. To use it in a project or job you attach it: the project receives it as an environment variable (QFS_AI_KEY by default) at deploy, restart or run. Nobody pastes it anywhere.

Because every place references the same key, Rotate reaches all of them at their next start, and Disable stops all of them at once. The key's Where it is used tab lists every attachment, and Requests shows its calls within a minute.

A project can also have its own key with one click (Settings > Resources > AI key > Issue the key), delivered as QCH_AI_KEY. Its spend then shows on the project's own Analytics tab.

Spend you can see at three levels

  • Per key: spend against the key's budget; calls stop when it is reached.

  • Per project: a project can get its own gateway team with its own models and budget, so a heavy project cannot use up another's allowance.

  • Per organisation: AI > Usage & budget shows tokens, estimated cost and requests for any period, split by key and by model, with Export CSV. An organisation admin can set a monthly budget with an alert at a percentage of it.

Keys can carry tags (for example product: pricing, tier: pro) to group spend in your own reports. Tags are labels only: they can never change a key's limits.

AI calls are paid in REM Credits from the organisation's wallet. Every plan includes a monthly amount, and Owners can top up from Plan and credits. The credits ledger lists every credit in and out, newest first. When the wallet is empty, paid AI pauses for that organisation until the next top-up or plan grant, and nothing else stops.

Bring your own provider key

An organisation that has its own contract with an AI provider can add its key under Connectors > AI providers. Its models then appear next to the platform's names, and its spend is tracked the same way. Organisations never see each other's keys, and a provider at a custom address waits for platform approval before it is used.

Get started

Create a key under AI > Keys, attach it to a project, and try a model in the Playground first. The AI guide has every field explained.

AIProduct

REM Credits: one ledger for every REM app

AI actions in REM apps are paid for in REM Credits. All of them are kept in one ledger, in REM Payments, with every movement and its reason. Here is how credits are granted, spent and taken back.

RREM Team18 Nov 2026 · 3 min read
One gateway for all your AI models, with spend per project · REM Platforms Blog