decision-provenance

Policy enforcement and execution infrastructure for AI agents. Merchval turns business rules into controlled, auditable actions using OPA, trusted evidence, approval workflows, and fail-closed decisions—helping teams deploy agentic AI safely across e-commerce, insurance, banking, and other high-stakes operations.

Machine-readable protocol for immutable, causally linked decision records with validator-backed integrity.

Three local-first demos that reconstruct operational trust from existing logs, receipts, and records. These are narrow reproducible slices of the broader Razorglint proof/governance architecture — designed to demonstrate trust continuity reconstruction, decision provenance tracing, and compliance drift detection under degraded environments.

Independent AI decision accountability for high-consequence decisions.

A protocol for recording and auditing how AI tasks traverse agents, models, providers, compute environments, and artifact-generation paths.

Early-alpha, simulation-only Python research framework for poker decision provenance, bounded CFR/CFR+ river experiments, synthetic leaks, and explanation-faithfulness checks; not a real-time bot.