Where this actually is
This is an experimental release. It is not production-ready, and it is not a preview of something nearly ready either.
We chose the lower of the two available words deliberately. Experimental means interfaces, defaults and configuration may change between releases without a migration path, and that nothing here has been run at length by anyone who did not write it. Run it in a lab. Read it. Break it, and tell us what broke. Do not put it in front of a fleet you would mind losing.
Production is not a property a project declares. It accrues after a thing has been run in many environments by people who did not write it. Saying otherwise before that has happened would ask you to take on trust exactly the thing this project exists to make verifiable.
So the site does not say enterprise-grade, battle-tested, or production-proven anywhere. What it says instead is what has been exercised, and where.
What is open, and what this release contains
There is no held-back edition. The product is open source — no paid tier, no build to unlock, no capability withheld from the published version. That is a decision about what this project is, not a description of a generous subset.
What the first release publishes is the enforcement path and the endpoint software: policy evaluation, TLS interception, browser-faithful egress, DNS control, east-west authorisation, mTLS device identity and admission, private-app publishing through the Connector, and the macOS and Windows agents. The operations layer — the admin console, multi-tenancy, high availability, multi-region — is being moved into the same public repository rather than kept out of it. That is a matter of sequence, and this page will say when it has landed.
In its default shape a decrypt-all deployment is two processes rather than one static binary, because the Edge's egress engine is cgo and links a native library. Describing it as a single binary would be the first thing an evaluator caught.
Experimental release, 2026-09-01.
The repository opens that day at github.com/lantern-networks/dsse-core.
What a licence is, and is not
With no vendor keys configured — the default, and what you get by cloning the repository — the software is unlimited. No signature is checked, no seat count is compared, and the licensing code never runs. There is no build to unlock and no feature withheld from the open version.
Where a signed licence is present, it declares a scope, and it may limit growth — refusing new enrolments beyond that scope. It may not stop steering, cut traffic that is already flowing, halt an edge, or disable features. Ending a commercial relationship must not take a running deployment down, and you can check the current state yourself: the predicates that would do it exist and have zero call sites outside their own tests.
What would have to be true before we use the word production
Written down now, so that the claim later has a defined meaning rather than being a marketing decision.
- Load and scale evidence beyond a single lab edge
- An operator running real traffic for people who are not us
- An independent security review of the interception and enforcement path
- Reproducible builds, so a released binary can be tied to the published source
- A support model with defined response expectations
Endpoint signing is partly done, and the part that is missing is the part that matters most on Windows. The macOS package is signed with a Developer ID and notarised. The Windows agent binaries are signed with an EV certificate. But the kernel-mode WFP driver still needs Microsoft attestation signing, which is applied for and not yet in place — so until it lands, loading the driver means putting the machine into test-signing mode. That is fine for a lab and disqualifying for anything else, which is why it is on this page rather than in a footnote.
When it does land, signed distribution is intended to be part of the open release rather than a commercially-gated artifact. Stated as intent, because it is not true yet.
The direction
The project is moving toward a single open codebase for the whole platform, rather than an open core with the operational layer held back. That is the intent and the work is under way; it is not a description of what this release publishes today. This page will say so when it is true, and not before.