Solutions · Enterprise agents
Authority for the agents you are already running.
Enterprise agents on Linux your company owns are the environment Vantio was built for and the one it runs in today. You can see the traffic, enforce policy on it, control the host underneath it, and keep the evidence.
This is the one node class where you can put every part of Vantio to work right now. Optics and Gate run on supported application paths, Phantom Engine runs on the Linux hosts you enroll, and Enterprise adds governance on top of those hosts.
What you are actually trying to do
You already have agents doing consequential work. They read from internal systems, call model APIs, move data between services, and in some cases spend money. The uncomfortable question is not whether the model is capable. It is what that process is permitted to reach, and how you would find out if it reached something else.
Most teams start from a blank spot. Nobody can name every agent that is running, nobody has a list of the destinations they use, and nobody can say which of them sit behind a control of any kind. Prompt instructions look like rules, but a prompt has never been a permission boundary.
So the work has a natural order. Get visibility before you change anything. Turn what you learn into policy you can preview on live traffic before it starts refusing things. Put a floor underneath the machine in case a process walks around the application path entirely. Then keep the record, because the person who has to answer for this in six months will not be in the room today.
What Vantio does on this node class today
Each product has one job, and you can adopt them in the order that matches how consequential the work is.
- Optics gives you the picture, free
- Wrap an agent process and Optics records where it connected, which process made the call, when, how much moved, and the trace that ties it together. It never stores prompts or completions, and it never blocks a request, so you can run it on production work while you are still deciding what your rules should be.
- How Optics works
- Gate turns that picture into decisions
- Once you know where traffic goes, Gate acts on it. You can refuse destinations, strip sensitive fields before they leave, cap request size and spend, and write the whole thing as policy in your repo. Dry-run shows you exactly what would have happened on real traffic before anything gets refused, and every decision leaves a record you can read. Gate is $499 a month after a 14-day trial, and trials are set up by our team rather than through a checkout page.
- How Gate works
- Phantom Engine holds the host when the application path is skipped
- Application controls only work when the agent actually uses them. On the Linux hosts you enroll, Phantom Engine watches TLS connections with eBPF, carries trace identity into forked children so a spawn is not a free pass, protects paths you declare, and scopes egress by cgroup and CIDR. Rogue Reconciliation is the part people care about most: when the host sees traffic that has no matching application record, it says so instead of staying quiet. It is $799 per enrolled Linux node.
- How Phantom Engine works
- Enterprise makes the proof durable
- When an auditor, a customer, or your own risk team asks who authorized something, Enterprise is the layer that answers. It keeps durable evidence and audit records on the same enrolled Linux hosts, puts dual-control in front of the dangerous moves, and gives the process a shape you can hand to someone else. Pricing is a conversation.
- How Enterprise works
A reasonable first month
You do not have to buy the whole ladder to get value out of the first rung. This is the sequence most teams end up following.
- 01
Wrap one agent
Pick the agent that worries you most and run it under Optics. Nothing changes about its behavior; you just stop guessing about its traffic.
- 02
Write the rules you wish you had
Turn the destinations and fields you saw into Gate policy, then leave it in dry-run until the preview stops surprising you.
- 03
Enroll the host underneath it
Put Phantom Engine on the Linux machine that agent runs on, so there is still a control point when something skips the wrap.
- 04
Decide what proof you owe
If the answer involves an auditor or a board, that is the moment Enterprise earns its place.
What is not available yet
These are the limits you would find out about in week two anyway, so you may as well have them now.
- No customer install has been validated by an independent operator. What we have proven so far, we proved on infrastructure we run ourselves, and that is a different thing.
- Optics and Gate cover the supported wrapped application path. A process that opens a raw socket, shells out to curl, or forks away can still egress without leaving an application-layer record. That is exactly the gap Phantom Engine is there to name.
- Gate does not mediate every browser, SaaS, socket, or tool path. The unsupported paths are written down in the compatibility matrix rather than glossed over.
- Phantom Engine applies only to the Linux hosts you enroll. Windows and macOS host enforcement are not available, and neither is control of every process on a machine.
- There is no self-serve checkout for Gate today. Trials are provisioned by the Vantio team.
- Vantio does not claim universal agent discovery. If an agent is running somewhere nobody told us about, nothing here will find it for you.
What stays true regardless
However far you take this, the ownership model does not shift.
- You own the policy. Vantio does not decide what your agents are for.
- You set the level of autonomy, and you can turn it down at any time.
- You can revoke a capability, and the enforcement point stops honoring it.
- You can export the evidence and take it with you.
- There is no hidden Vantio remote override on your infrastructure.
- Independent safety systems keep the authority they were designed with. Vantio does not replace an interlock, an e-stop, or a local controller.
Start where it costs you nothing
Optics installs from npm or PyPI and does not block a single request, so the first step is genuinely low risk. If you would rather walk through your hosts, your agents, and where enforcement should begin, we are happy to do that first.
Other node classes are covered honestly on Solutions, including the ones that are still design work.