A productized connectivity layer for AI and SaaS operators.
If your platform needs to reach customer data on customer infrastructure, you have three options. Build it yourself. Push the work to your customers via VPN setup guides. Or run Ewii. This page is for operators considering option three.
FIG. 1Hub binary anatomy — five internal subsystems from trust registry to metrics emission.
Hub deployment characteristics
The Hub runs as an OCI container in your infrastructure: your VPC, your Kubernetes
cluster, your bare-metal. The control plane is a single binary; a connection-handling
relay node accepts up to 10,000 concurrent Client connections by documented default,
and a Hub region scales horizontally with N relay nodes behind a Layer-4 load
balancer. We provide the binaries, the images signed with Cosign, the deployment
manifest, the capacity-planning notes, and the operational runbook. Your team owns the
rollout. We back you up.
Multi-tenancy at the Hub level
A single Hub serves many customers concurrently. Tenants are cryptographically isolated:
each customer’s SPIFFE trust bundle is independent, and the Hub enforces tenant boundaries
at the connection layer before any application traffic flows. There is no shared-key
cross-tenant code path. A compromise of one tenant’s Client provides no path to another
tenant’s traffic via the cryptographic boundary.
White-label and rebranding options
The Client container can be retagged and branded with your product name. The web
experience your customers see when installing the Client — setup instructions, status
page, support links — is fully customizable. Your customers see your brand; we sit behind
it. Trust attribution stays with us only when a customer asks where the security model
comes from, and you point them here.
Operational fit
Prometheus metrics out of the box: connection counts, handshake latency, ChaCha20-Poly1305
throughput, replay-window rejections, certificate-rotation events. Structured logs (JSON)
for SIEM ingestion. Indicative SLA targets are 99.95% steady-state availability per Hub
region and 99.9% during planned windows; contractual SLAs are negotiated per engagement
and depend on your chosen deployment topology and the support tier you select.
PagerDuty integration is your call; we don’t ship a paging system.
The Architecture Review process
Every operator engagement starts with a 60-minute Architecture Review. Our senior
architect and security lead attend. No account executive, no sales pitch. We walk through
your platform’s data flow, your customers’ security posture, and the deployment shape that
fits both. You leave with three artifacts: a draft architecture diagram, a
security-responsibilities matrix (who owns what), and a deployment plan. Whether you adopt
Ewii after that is your call — we don’t bill for the review.
