Magenta Canon is a private-access proof system in controlled evaluation. There is no public self-serve pricing, no public package to install, and no licensing tier you select. Evaluation is a granted, private evaluator path for design partners: you run the proof loop yourself and review what the system proves. Future licensing is a separate conversation, reviewed on its own terms after evaluation.

  • No self-serve signup and no public install path.
  • Access is granted deliberately, one evaluator at a time.
  • Licensing terms are not published and are not a tier selected here.

EVALUATION ACCESS

Private evaluation, not a plan you buy

Magenta Canon is a private-access proof system. Evaluation is a controlled evaluator path — you review what the system proves, not a self-serve plan you purchase.

How evaluation works

Evaluation begins with a granted, private checkout, scoped to design partners. You run the proof loop yourself and see how gate, witness, receipt, and verify fit together. There is no public self-serve signup and no public package to install — access is granted deliberately, one evaluator at a time.

Future licensing

Any future licensing is a separate conversation, reviewed on its own terms after evaluation. It is not a tier you select here, and no pricing, packaging, or licensing terms are published on this page.

What this page does not claim

No self-serve plan, no public package, no compliance certification, and no production-SaaS offering. Magenta Canon is a reference implementation under private evaluation. Its production multi-tenant control plane is partly merged and not production-active, and hosted operation is not built — none of it is available to buy.


What an evaluator gets

A granted private checkout and the runnable proof loop: an allowed action, a blocked action, the downstream tool's own log showing the blocked call never arrived, an independent verification, and a tampered bundle that fails. Then a technical review of the gateway and verifier source, a diligence packet on request, and a working session against one real workflow of yours where a wrong agent action would hurt.