Sovereign Inference (JADRO)

The brain that never leaves your environment.

SPNT's agentic modules — investigation, offense, and agent control — reason on a large language model. JADRO is the layer that decides where that reasoning runs, so you never have to trade intelligence for sovereignty.

Two modes, one platform

• Cloud reasoning. By default, reasoning runs on managed frontier models (Claude) for maximum capability, with no customer data used for training.
• Sovereign reasoning. For regulated, classified, or air-gapped environments, JADRO runs reasoning on a local model inside your own infrastructure — zero egress, no third-party API calls.

The same modules, prompts, and graph queries work in both modes. Moving from cloud to sovereign is a deployment decision, not a rebuild.

Why it matters

• Data residency. Your security graph contains your most sensitive operational data. With JADRO on-prem, that data and the prompts built from it never leave your boundary.
• Air-gapped operation. Defense, government, and critical infrastructure can run the full agentic platform with no internet connectivity.
• Sovereignty by design. SPNT is a European company in North Macedonia — outside the EU and not a US entity. JADRO lets you keep inference under your own jurisdiction and control.

What runs on JADRO

• ISTRAGA — autonomous investigation and adversarial reasoning.
• NAPAD — agentic offense planning and exploit reasoning.
• ZASTAPNIK — agent and MCP reasoning, kept local so tool context never leaves.

Integration with other modules

JADRO is the substrate's reasoning engine. Agents in ZASTAPNIK act on the graph; the agentic modules reason on JADRO. Because reasoning reads directly from the same substrate, analysis is always grounded in current, validated evidence rather than stale snapshots.