Overview
Proxility connects networks that should not trust each other. You install a connector in each network, write policy about identities, and Proxility carries the traffic — mutually authenticated, encrypted end to end, and denied by default.
The problem it replaces#
The usual way to reach a customer's private estate is a site-to-site VPN. That gives you a tunnel between two subnets and, with it, a set of problems:
- Trust is all-or-nothing. Once the tunnel is up, anything on one side can reach anything routable on the other. Segmentation becomes a firewall exercise, maintained by hand, drifting from intent.
- Addresses are the unit of policy. Rules are written about CIDR ranges. Re-address a subnet, autoscale a workload, and the rules quietly stop meaning what they meant.
- Inbound exposure. Someone has to listen. That is a public endpoint, a firewall change request, and an attack surface.
- Overlapping RFC1918 space. Two estates both using
10.0.0.0/8turn into a NAT puzzle. - Credentials that outlive their purpose. Pre-shared keys and long-lived certificates sit in configuration for years.
What Proxility does instead#
| Concern | Proxility |
|---|---|
| Reachability | A named service, published explicitly by one connector |
| Unit of policy | Workload identity, not IP address |
| Direction of setup | Outbound only, from both sides |
| Credential lifetime | 24-hour certificates, auto-renewed |
| Overlapping subnets | Irrelevant — services are named, not routed |
| Default posture | Deny |
A connector publishes payments.prod.acme for one TCP port in its network.
Another connector's workload resolves that name and connects. Policy decides
whether the session happens; the two connectors negotiate keys directly; the
edge relays bytes it cannot read.
What you get#
- Connectors for Linux (deb, rpm, container), Windows and the major cloud marketplaces, plus a Kubernetes deployment for in-cluster use.
- A console for enrolment, policy, service catalogue and the decision log.
- A control API and CLI so every console action is automatable.
- Regional edge at the ScaiLabs standard region/level naming:
nlwest-stdandnlsouth-stdat launch, both in the Netherlands.
When to use it#
Use Proxility when you need a specific service reachable across an organisational boundary and you want that reachability to be explicit, identity-based and logged. It is a good fit for reaching customer estates from a SaaS control plane, for joining cloud accounts without peering, and for giving a partner or contractor access to exactly one internal application.
When not to use it#
- You need full IP routing between two networks. Proxility publishes services; it is not a router. If you genuinely need arbitrary L3 reachability — multicast, dynamic routing protocols, broadcast discovery — a VPN or a cloud transit gateway is the right tool.
- Latency-critical traffic inside one data centre. Two workloads in the same broadcast domain should talk directly.
- You cannot run software in the network. Proxility needs a connector. There is no appliance-free mode.
Next#
Quickstart joins your first two networks in about fifteen minutes.