Security model
Six questions security teams ask us.
Every good security team asks them. Here are straight answers, based on how LayerV works today.
- 01What does someone actually reach?Only your app, at its private address, by way of LayerV.
- 02What does a scan find?Nothing of yours. No address, no port, no reply.
- 03Doesn’t that make LayerV the target?One hardened edge we defend, instead of every app you’d have to.
- 04What if LayerV is breached or down?You hold the off switch. Access starts and ends on your side.
- 05Why not run OpenNHP ourselves?Because LayerV is OpenNHP, finished, run by the people who wrote it.
- 06Can LayerV replace our other security tools?For app access, it can. But we believe in depth: LayerV is the invisible first layer.
When someone gets through, what are they actually connected to?
Only your app, at its private address, by way of LayerV.
What answers is your own service, exactly where it runs today: a dashboard on an internal server, an API in your AWS VPC, an MCP server, or an app on a laptop. Only the connector beside it ever opens its private address, such as 10.0.1.5:8080 or localhost:3000. On AWS, the connector runs as a container in the same VPC and reaches the service on its private IP.
Why that’s safe: the connector starts both of its connections from inside your network, the way a laptop reaches a website. First it sends one encrypted, authenticated knock over UDP 443, then it opens an encrypted tunnel over TCP. Your firewall lets them out and never has to let anything in. With nothing listening, there is nothing on your side for the internet to connect to.
When a person or AI agent opens a link, LayerV checks their grant on every request before anything enters the tunnel. Like any access gateway, it does that inside its own hardened edge and stores nothing that passes through.
If I scan it or look it up in DNS, what do I see?
Nothing of yours. No address, no port, no reply.
Your app has no public address, no open port, no listener, no banner and no certificate to read. The connector only ever connects outward.
What a scan does find is LayerV, the way it finds your VPN or identity provider today. That’s the design: one well-defended front door that isn’t yours to patch, instead of every app on display.
One step on your side: close any old public route to the app, such as a legacy hostname or load balancer. LayerV can’t hide a door you leave open.
Your servers are on the internet. Doesn’t that make you the target?
One hardened edge we defend, instead of every app you’d have to.
Today every app you run is its own target, with its own ports, its own login page and its own patch schedule. LayerV collapses all of that into one single-purpose edge that we build, run and defend full-time.
- The access edge drops traffic from anyone who hasn’t presented a valid cryptographic knock. No reply, no banner.
- The edge limits the rate of knocks from each source.
- The connector endpoint, connect.layerv.ai, admits only a connector holding a short-lived token LayerV issued to that exact connector. Verified on our production edge: a connection without a valid token is refused in under a second, before any session exists.
Even knowing every detail on this page, an attacker gets no closer to your app.
What you defend today, for every app
- An open port
- A login page anyone can try
- A TLS certificate in public logs
- Patches before attackers find the gap
What you defend with LayerV
Nothing on the public internet.
Your apps have no public front. LayerV defends its edge, full-time.
Are we just putting our trust in you? What if you’re breached or down?
You hold the off switch. Access starts and ends on your side.
The tunnel starts on your side. Stop the connector to close the path for everyone. Revoke a link to remove the access it grants.
- Every grant covers one app, is short-lived, can be single-use or tied to an email address, and can be revoked in a click.
- LayerV adds a lock in front of your app. Your app’s own sign-in stays exactly where it is.
- The key that issues access lives in AWS KMS hardware, never leaves it, and every use is logged.
- If the access edge goes down, requests through it fail. The outage does not open a public route to your app.
Built in the open: the protocol, OpenNHP, and our client software, the connector, CLI and SDKs, are open source, so your team can inspect what runs in your network. A third-party penetration test and SOC 2 are planned for 2027.
Your off switches
- Stop the connectorThe tunnel starts on your side. Stop it and nothing can reach the app.
- Revoke a linkOne click, and that person or agent is out.
- Let it expireEvery grant has a lifetime. Single-use links close after one open.
What’s unique to LayerV? Why not just run OpenNHP ourselves?
Because LayerV is OpenNHP, finished, run by the people who wrote it.
LayerV’s founders wrote OpenNHP. Our CEO, Justin Posey, and CTO, Ben Chen, are the authors of its IETF draft, and our team co-authored the Cloud Security Alliance specification. Nobody knows the protocol better, and we keep LayerV current with it, so you never need an NHP expert on staff.
The protocol hides a resource. LayerV turns that into a product your whole company can use, with everything a production deployment needs built in:
- A managed, always-on edge: keys, grants and revocation run for you, and your app connects with one container.
- Access per app, per person, per agent: grants that expire, work once, or require a verified email.
- Nothing to install for the people you let in: they open a link in any browser.
- Built for AI agents as well as people: a CLI, SDKs, a REST API and an MCP server.
- Deploys where you run: Docker, Kubernetes and ECS.
You can run OpenNHP yourself, and its openness is why you can trust ours. LayerV is what you’d otherwise have to build and run yourself, ready today.
We already run Tailscale, Appgate, CrowdStrike and more. Can LayerV replace them, or why add it?
For app access, it can. But we believe in depth: LayerV is the invisible first layer.
For your web apps, APIs and AI tools, LayerV can stand on its own: no VPN client, no broker, no public login page. Teams can move contractors, partners and AI agents off the VPN entirely.
You don’t have to rip anything out, and we don’t think you should. Endpoint tools like CrowdStrike watch your devices. Your identity provider decides who someone is. Your zero-trust tools manage your network. LayerV adds the layer none of them provide: it makes your apps invisible, so attackers and their AI never reach them in the first place.
- For people outside your network: contractors, partners and clients open one app in a browser, with no client, account or network enrollment.
- For AI agents: give an agent one app for one task, then take it back.
- Down to one app: a grant reaches a single app, not a network segment, and can expire or be used once.
- Alongside what you run: your identity provider, endpoint security and network stay as they are.
Bring us your hardest question.
We’ll walk your security team through the architecture on a call, and answer anything this page doesn’t.