How It Works

Step by step

01

A request arrives

A request enters MNDe before runtime. The request is treated as input to a decision system, not as a best effort hint.

02

Preflight normalizes the input

Preflight reduces the request to a canonical form. Unknown fields and malformed structures are rejected so policy sees the exact same shape every time.

03

ARM calculates limits and cost

ARM computes cost, scope, and declared limits before execution. If the request exceeds policy or capacity, it is stopped before resources are consumed.

04

Policy is evaluated

The normalized request and calculated bounds are evaluated against policy. There is one policy outcome for one input and one policy state.

05

RAM0NA returns allow or refuse

RAM0NA returns the decision as a deterministic result. Allow or refuse. There is no silent downgrade, no warning mode, and no hidden branch.

06

Decision is signed and stored

The final result is signed and written as a replayable record. The same request and policy can be re-run later and verified against the stored outcome.