AI features that stay up and in budget
The problems it removes
- A provider outage takes your feature down
- Failover moves requests to the next healthy provider or model.
- One customer runs up a bill
- Give each feature or tenant its own key with a monthly budget and model rules.
- Nobody knows what a feature costs
- Every request is costed; tag requests with the X-Title header and use a project per feature.
- User input reaches the model unfiltered
- Guardrails redact personal data and block prompt injection before the provider sees it.

In 1953 an operator put each call through by hand. DigitSpace routes every request.
Switchboard operators, NIH Clinical Center, 1953. NIH History Office, public domain, via Wikimedia Commons.
A common set-up
| Feature | Key | Rules |
|---|---|---|
| Support reply drafts | support-drafts | auto:economy, $300 a month, guardrails redact PII |
| Document Q&A | doc-qa | claude-sonnet-5-5, $800 a month, Anthropic and OpenAI only |
| Internal tools | internal | Any model, $100 a month, office IP range only |
An example; names and numbers are yours to choose.
Questions
Can we create keys from our own backend?
Yes. Master API keys manage projects and keys for your organisation over an API, so each tenant can get its own key and budget.
Does the gateway add latency?
Routing happens inside the gateway with no extra round trip; you mostly see the provider's own speed.
Can we keep prompts out of your logs?
Yes. Retention is off unless your organisation turns it on.
Read next
Live in a minute.
Create a key, change the base URL, send a request. No card needed with your own provider keys.
Start free