When the work disappears, so does your cost of goods.
The agents you were reselling at a thin markup, gone from the endpoint. The labor that ate your margin, mostly gone with it. Your cost to deliver a seat drops hard, and the spread between what you bill and what it costs you widens.
Every agent on a Windows seat exists because the OS is persistent.
On Scylos the entire agent stack is eliminated, only the off-device layers remain: email security and cloud/SaaS backup, plus the flat-fee substrate and Switchboard. The comparison below is built on a representative SMB managed seat of ~50 seats per account, run at the ~50–60% industry gross margin.
The agent stack, eliminated
Where the labor goes
Truck rolls are modeled as a small residual on Scylos ZeroCore, not zero: most crashes, malware, and drift reset remotely with no dispatch, but a genuinely dead box still needs hands to insert the USB. We assume no reduction in hardware-failure frequency. Identity, network, and how-to tickets don't shrink, the machine itself just stops generating tickets.
More margin per seat, every month
Across a full account that compounds into materially higher margin every year. Or pass it on: cut the client's bill to win the deal and your margin still rises.
“Client pays less, MSP earns more, the offer an agent-based stack structurally cannot make.”
About the economics
Are these your numbers or mine?+
These are representative, sourced from published wholesale ranges and ticket-handle benchmarks. Specific figures belong in the calculator, drop in your own seat price, count, and ticket mix and it personalizes the result.
What stays in the stack?+
Identity, email security, and cloud/SaaS backup. Scylos removes the endpoint that generated most of your tickets; it doesn't replace your identity or network layers.
Why isn't labor zero?+
Because identity, network, and how-to tickets are about people and accounts, not the machine. Those persist. Only the OS-, malware-, drift-, and hardware-driven labor collapses.
See the stateless endpoint on your own hardware.
Flash an idle machine into a live endpoint and run your real workloads. You buy no hardware and sign nothing.