Skip to main content
hiloop can run as hosted SaaS, hybrid, BYOC, or on-prem / air-gapped deployments. The product model is the same: requests are authorized at the edge, backend services use the stamped identity, runtime resources execute workloads, and telemetry is stored under organization scope. The boundary changes based on who operates the infrastructure and where data is stored.

Boundary summary

Platform telemetry is hiloop’s own service telemetry, separate from the run telemetry your agents produce. Its spans carry a fixed allowlist of low-cardinality identifiers and nothing else: request bodies, payloads, and secrets are never attached in the first place, rather than attached and then stripped. Your captured run telemetry is governed by redaction instead.

Identity boundary

Every north-south request is checked before it reaches a backend. The edge resolves an API key or session to an organization, user, and scope, then forwards trusted identity context to services. Services derive organization scope from that context, not from untrusted request bodies. Interactive shell sessions cross the same boundary and add no second one. hiloop sandbox ssh carries the same organization-scoped API credential as every other request, and the API authorizes each connect against the organization that owns the sandbox before it moves any bytes. There is no separate SSH credential, no jump host, and nothing for you to rotate. See shell access.

Data boundary

Large payloads matter most for trust reviews. Hosted deployments store event and workspace payloads in hiloop-managed storage. Hybrid deployments keep selected payload storage under customer-owned object storage. On-prem and air-gapped deployments keep the data plane inside the customer environment.

Operational boundary

Hosted deployments optimize for speed and managed operations. On-prem deployments shift more responsibility to the customer: Kubernetes, PostgreSQL, object storage, registry, TLS certificates, backup/restore, and local observability. BYOC and hybrid sit between those ends. On-prem, hiloop operates nothing in your environment and holds no access into it — see hosted and self-managed.

Verify it yourself

The boundaries above are observable, not just asserted. On an on-prem deployment you can confirm:
  • The data boundary holds. Point the deployment at your own PostgreSQL and object storage, then confirm control-plane metadata and event/workspace payloads land only there.
  • The air-gapped posture is real. Disable support telemetry (supportTelemetry.enabled=false) and confirm no service pod can reach the public internet — the network requirements check exercises this directly. In the default supported posture, the only outbound flow is a fixed platform-health metrics contract; workload data and event payloads never leave your environment in either posture.
  • The identity boundary holds. Every backend derives tenancy from edge-stamped identity, not from request bodies — the threat model has the concrete egress, secret, signing, and non-root checks.