Deployment options
Where the API Platform can run, from Safeguard's hosted API to an air-gapped install, how mature each option is, and which models each one can use.
Early access. The API Platform is deployed in Safeguard's hosted service but switched off by default. Safeguard switches it on per workspace.
| Option | What it involves | Maturity |
|---|---|---|
| Safeguard API (hosted) | Safeguard runs the gateway. Your code calls https://api.safeguard.sh/gpt/v1 with an API key from any OpenAI-compatible client, and each request goes to the provider of the model you choose. There is nothing to install or operate. | Early access |
| Public cloud | Safeguard deployed on the public cloud you choose. The platform is cloud-agnostic, so it isn't tied to one provider. | GA |
| Private cloud (VPC) | A single-tenant deployment in a private network, isolated for one customer. For Enterprise. | GA |
| On-prem | The full platform in your own datacenter. | GA |
| Air-gapped | The complete platform with no outbound connection, including Safeguard's own models. Updates arrive as signed offline bundles. | GA |
The GA labels describe running the Safeguard platform in that environment. The API Platform itself is in early access in every option, switched off until Safeguard turns it on. For any deployment other than the hosted API, ask Safeguard to scope the API Platform with you.
Buying through your cloud provider's marketplace is not available yet for the API Platform. See Procurement.
Which models can run where
On the Safeguard platform, Safeguard's own models (Griffin, Eagle, Lion and Zero) and open-weight models can run inside your own environment: in a public or private cloud, on-prem or air-gapped.
Safeguard's own models are available in 245 countries and territories (the ISO 3166-1 list minus the exclusions). They are not offered in Cuba, Iran, North Korea or Syria, nor in the Crimea region of Ukraine or the so-called Donetsk People's Republic (DNR) and Luhansk People's Republic (LNR) regions of Ukraine. See Safeguard models.
Through the API Platform, in this version:
- every catalog model is served by its provider, or by the host the catalog names for an open-weight model, over the internet;
- Safeguard's own models cannot be called yet, in any option. See Safeguard models.
Closed models, such as those from OpenAI, Anthropic, Google and xAI, run only on their providers' infrastructure. They are never installed on-prem or in an air-gapped environment, so using one always means sending the request to its provider's API.
So in an air-gapped deployment the API Platform has no model it can serve yet.
The hosted API
- An OpenAI-compatible API at
https://api.safeguard.sh/gpt/v1:POST /chat/completions,GET /models,GET /models/{id}andGET /key. - Requests are authenticated with an
sg-api-key. See Security and compliance. - Usage is billed at the provider's list price, with no markup, from prepaid credit. See Pricing and billing.
- Your prompts go to the provider of the model you choose. See Privacy and data and Regions and data residency.
Getting a dedicated or self-hosted deployment
For committed volume or a dedicated deployment, email hi@safeguard.sh, and see Procurement.
Related
Security and compliance
How the API Platform protects requests and keys: TLS in transit, hashed single-purpose API keys, checks that fail closed, an audit trail, and where to find Safeguard's compliance reports.
Procurement
How to buy the API Platform: directly from Safeguard, under an enterprise agreement, or through your cloud provider, and what your vendor review will need.