Knowledge ships in three deployment shapes. Pick based on data residency requirements, compliance posture, and operational appetite.
SaaS (Asplenz-hosted)
Where it runs : Asplenz-managed infrastructure in a chosen region.
Who operates it : Asplenz SRE.
What you get :
- Fastest time-to-first-decision. Provisioning is minutes, not weeks.
- Rolling upgrades, patches, and monitoring included.
- Certifications trail (design-partner tier today ; SOC 2 + ISO 27001 as the cohort matures).
What you give up :
- Data physically leaves your perimeter (SaaS provider = Asplenz).
- Certification posture is Asplenz's, not yours ; audit responses cite Asplenz controls.
Suitable for : POCs, teams without compliance blockers on external SaaS, verticals where the data is not itself regulated (agent tool orchestration, internal ops).
Isolation : row-level multi-tenant by default. Per-tenant infrastructure available for enterprise plans.
Private cloud / VPC
Where it runs : your cloud account (AWS, Azure, GCP, ...).
Who operates it : jointly. Asplenz provides the deployment tooling ; you own the account and its IAM.
What you get :
- Data never leaves your VPC.
- You control network policies, IAM, security groups, and audit logs.
- Compliance posture inherits your controls.
What you give up :
- Operational overhead : you patch, upgrade, monitor.
- Rolling releases require your acknowledgement.
Suitable for : regulated verticals (banking, healthcare) where residency + control are contractual. Any org with mature cloud practices.
Provisioning : Asplenz ships Terraform modules for AWS + Azure + GCP. Deployment takes hours to days depending on your review cycles.
Data plane vs control plane : your VPC runs both. No management plane call-home.
On-premise
Where it runs : infrastructure you operate (bare metal, private data centre, air-gapped environment).
Who operates it : you.
What you get :
- Complete control. Data never leaves your infrastructure.
- Air-gap possible (no network dependency on Asplenz).
What you give up :
- Full operational overhead. You handle everything.
- Support cycles are longer (fewer telemetry signals to Asplenz).
Suitable for : air-gapped environments (defence, restricted networks) ; jurisdictions with strict on-prem-only requirements.
Runtime dependencies :
- Postgres 15+ (managed or self-hosted).
- Your LLM provider (Anthropic, OpenAI, ...) if you enable
knowledge-aifor verdict prose. Otherwise no LLM dependency.
Everything else : bundled in the Knowledge distribution.
Shape comparison
| Dimension | SaaS | VPC | On-prem |
|---|---|---|---|
| Time to first decision | Minutes | Hours-days | Days-weeks |
| Data residency | Asplenz region | Your VPC | Your infra |
| Certification inheritance | Asplenz's | Yours | Yours |
| Operational overhead | None | Moderate | Full |
| Upgrade cadence | Rolling (Asplenz) | You approve | You control |
| Air-gap possible | No | No | Yes |
| Per-tenant isolation | Row-level default ; per-tenant available | Row-level default ; per-tenant available | Row-level default ; per-tenant available |
What is the same across shapes
- API surface : identical. Callers do not know which shape they hit.
- Signed verdicts + JWKS : identical envelope, per-tenant kid.
- Audit surface : identical Consultation shape.
- Enforcement model : PEP does the same work regardless of shape.
The difference is where the servers live and who operates them - not what the software does.
Getting started
- Read Security - Asplenz's current controls.
- Read Keys inventory - what's in the deployment.
- Talk to us with your compliance / procurement requirements ; we scope the shape and timeline together.
Related
- Multi-tenant setup - operating multiple tenants on one deployment.
- Compliance posture - certifications + design-partner tier.
- Keys inventory - key storage per shape.