What teams tell us makes them pick it.

01

You choose how hard to protect it

Two integration models, your call, per feature. The simple path is a few lines — verify the licence, branch on the result. Its honest limit: that is one boolean gating a branch, and a determined attacker can flip it with a one-instruction binary patch in your own executable. The load-bearing path removes the branch entirely — the licence physically carries an encrypted data key, and a real feature asset is sealed under it and opened at runtime via the LicenseAsset type. A forged or patched licence yields the wrong key, so the feature just produces garbage; there is no verdict to patch. It ships as a toolchain — rg_bind_asset seals an asset at build time, a CMake helper wires it in, rg_bind_lint checks your binding — and it is incremental: protect one high-value feature, leave the rest on the simple path.

02

Signed file format you can audit

RockyGuard licenses are signed JSON. The payload is plain text — your security team can read it, diff it, and verify it with any standard Ed25519 / RSA-SHA256 implementation. No proprietary blob format, no opaque binary container, no need to trust a vendor tool just to inspect what is inside a license.

03

Public threat model

We publish the nine-row threat model on the public website verbatim from the customer docs. Every concrete attack named, every defense mapped, every residual risk acknowledged — including the "memory dumping is out of scope" row. Most products in this category bury this material; transparency is the differentiator.

04

No SaaS dependency, no phone-home

License verification runs entirely offline against a static public key. The optional floating server runs on your customer's LAN — RockyGuard never sees it, never logs it, never knows you have a customer at all. The only outbound network call the library makes is anonymous HTTPS time-anchor verification to a rotating pool of public hosts; no telemetry, no activity beacons.

05

Lightweight integration

A small C++17 library and a handful of CLI tools. No agent to install on customer machines. No daemon to manage. No SDK to register. The Quick Start integration is a single function — three library calls inside one try/catch. That is the floor, not the ceiling: when a feature has to survive a determined attacker, the same library seals it under the licence instead of gating it behind a branch (see the protection-choice reason above). Static or shared library, your choice. Windows and Linux today, macOS in v1.3.

06

Honest scope

RockyGuard is C++17 desktop and server applications, Windows and Linux x64, glibc 2.34 or newer. We list this on the landing page deliberately. Buying licensing infrastructure and discovering a platform gap two weeks in is the worst possible experience for both sides; we would rather lose a sale at the door than ship a customer into a corner.

07

Customer-onboarding transparency

The Phase 4.1 customer-onboarding dry-run report — three sub-agent runs against the shipped customer zip with every gap they found, every fix made in-session, and every gap deferred — is shipped in the docs and published openly. You can read what a fresh evaluator hit before you ever start your own evaluation, and what we changed because they hit it.