Deterministic governance architecture for systems that operate on evidence, not trust. 16 specs covering Bitcoin custody, AI governance, embodied agents, and autonomous agent security. Post-quantum. Fail-closed. Authority-separated.

External kill switch for autonomous runtimes. Validate at enforcement boundaries. Revoke to halt execution.

Fail-closed execution guard for workers, services, and agents

完全ローカル×思考可視化×三権分立の軽量AIコア。ネット/外部DBなし。state保存/復元、4D意味距離(geo/strict)、署名付き役割カード、𝒢実装。

Open, verifiable agent framework — fail-closed orchestration, evidence-based verification, tamper-evident audit chains (SHA-256/HMAC).

Cite or shut up. LLM answers without verifiable sources are liabilities. Extractive RAG with line-anchored citations that fails closed when confidence is low. Deterministic, no API keys.

Official Python SDK for MachineID - simple device registration and validation for AI agents.

Fail-closed enforcement kernel for AI agents. Structural guardrails, mandatory choke points, exception-based blocking, and regression-backed boundary protection.