DeepIntShield Cloud
Fully hosted at app.deepintshield.com. Nothing to deploy - sign up, create a virtual key, and point your SDK at the gateway. Best for most teams.
DeepIntShield offers three deployment shapes. They share the same gateway and APIs; what changes is where your data lives, who operates the infrastructure, and the latency and availability envelope.
DeepIntShield Cloud
Fully hosted at app.deepintshield.com. Nothing to deploy - sign up, create a virtual key, and point your SDK at the gateway. Best for most teams.
Enterprise VPC
The data plane runs in your own cloud (GKE / AKS / EKS); DeepIntShield manages the control plane. No prompt, response, key, or audit data leaves your VPC. Best for strict data-residency and compliance requirements.
Self-Hosted (Enterprise)
Run everything - control plane and data plane - in your own infrastructure, including fully air-gapped. Delivered with our solutions team. Talk to sales to scope it.
| Cloud | Enterprise VPC | Self-Hosted | |
|---|---|---|---|
| Where your data lives | DeepIntShield Cloud | Your cloud account | Your infrastructure |
| Who operates it | DeepIntShield | DeepIntShield (control plane) + you (data plane) | You |
| Get started | Minutes - sign up | Guided onboarding with our team | Talk to sales |
| Network | Use app.deepintshield.com | Outbound-only to the control plane; no public ingress | None required (air-gap option) |
| Best for | Most teams; fastest path | Data residency, regulated industries | Air-gapped, on-prem, full control |
Hosted and operated by DeepIntShield. There is nothing to install - see the Gateway quickstart:
https://app.deepintshield.com.You host the DeepIntShield data plane (gateway + guardrails + ML detectors,
with your own Postgres, Redis, and vector store) in your GKE / AKS / EKS
cluster. DeepIntShield runs the control plane at app.deepintshield.com for
policy authoring, dashboards, and billing.
What stays in your VPC: every prompt, completion, embedding, retrieved chunk, provider key, and audit record. Only signed configuration flows down and aggregate counts flow up, over a single outbound-only, pinned mutual-TLS connection - there is no inbound path into your environment, and the control plane is not on the ordinary request hot path. Local policy enforcement, configuration-cache misses, instrumentation, and upstream calls still add workload-dependent latency; benchmark the complete VPC path under expected concurrency and failure modes.
See Enterprise VPC deployment for the step-by-step install.
Run the full stack - control plane and data plane - entirely in your own infrastructure, including fully air-gapped environments. Self-hosting is an Enterprise offering delivered with our solutions team, who provide the deployment bundle, license, and a guided rollout for your platform (GKE, EKS, AKS, on-premise, or air-gapped).
SDK and Server 3.2.0 support server-managed LangGraph onboarding when your operator enables it on the gateway and supplies the isolated worker runtime. The operator can enable it for every workspace at once; limiting it to selected workspaces is optional. Ordinary gateway hosting does not enable that worker. The default runner requires an operator-managed Docker host, immutable images, shared staging storage, enforced egress policy, and HTTPS gateway access. As an optional deployment alternative, the gateway can send each run over HTTPS to a separately deployed worker service that runs the same pinned worker image; the operator must restrict that service to the gateway and give every run a fresh, isolated instance. Native Cloud Run or Kubernetes hosting of the gateway alone does not provide either runtime.
Confirm worker availability with your service operator and that the worker image contains every package your program imports. The worker runs its own SDK, Python and package releases; your client’s versions are recorded as evidence and never need to match. Local Agentic execution remains the default.