Trusted Partners
Seamlessly integrate with leading platforms
Where SecureAI Guard sits in your stack
SecureAI Guard runs on the path between your application and the model it calls. Every control is applied server-side, to the request on the way out and to the completion on the way back, which means the integration point is your own application code rather than the model provider's console. That is the property that makes the list of platforms above meaningful: the guard is inspecting a request your application assembled, before it leaves your network.
Enforcing an explicit allow-list of models and providers on every call is also where the model inventory comes from. The set of providers your organisation is actually using is derived from traffic rather than from a spreadsheet somebody maintained for a quarter, and an unreviewed provider appearing in that traffic is both a security finding and a data-transfer finding. What each control inspects and enforces sets out all six, including what each one does not do.
What is not published yet. There is no per-provider compatibility matrix — no page that says which model providers, SDK versions and deployment shapes are supported and tested. Until there is, treat the platforms above as the environments SecureAI Guard is built to run in, not as a tested support list, and ask us about your specific stack before you plan around it.
How it gets deployed
Two shapes, and which one you need is usually decided by whether prompt content is allowed to leave your network.
- Cloud marketplace listings — procurement through a cloud provider's marketplace, so the spend goes against an existing commitment. The current status of each listing is maintained on that page, per listing, and is deliberately not restated here: one status claim in one place is the only way it stays true.
- Self-hosted and air-gapped deployment — SecureAI Guard inside your own network or VPC, for workloads where prompt content cannot leave your infrastructure at all.
Integrating it
The integration is application-side and takes four steps: get access to the package, configure the guard for your environment, wrap the model call on both the input and the output side, and turn on monitoring and the audit trail. The integration walkthrough has the middleware, API-route and monitoring code for each.
Deciding whether you need this layer at all
The platforms above all ship security controls of their own, and for some applications those are enough. Rather than argue that here, the comparisons set out how the options differ and how to measure them against your own traffic — including the case for building the layer yourself.
- LLM security alternatives, and how to compare them — the open-source and platform options, described from their own documentation and linked to it.
- Build vs buy: writing your own LLM guardrails
- How to benchmark an LLM guardrail on your own traffic