llm-governance

企业级 AI 资源纳管平台，提供统一 AI网关、Token调度能力，纳管 OpenAI、Azure、Claude、DeepSeek 等主流模型，并支持 MCP 工具与 Skill 的集中注册分发。具备内外双轨定价、成本归因、统一身份认证、安全审计与效能报表，帮助企业精准控制成本、量化 ROI，高效治理 AI 资产。

A disciplined methodology for AI-assisted software development. Covers architectural constraints, validation hooks, session governance, and PAG (Pattern Abstract Grammar) for structured AI collaboration. Copy claude-setup/ into your project to start.

Control plane plugin for Hermes Agent — pool many free/cheap model providers + local AI CLIs, switch brains from Telegram in one command (even while the model is down), delegate work cap-aware with auto-fallback. Run your agent at ~$0 without hitting quota walls.

Five Tests Standard (5TS): a vendor-neutral published standard for verifiable AI governance through proof-carrying decisions. Includes schemas, machine-checkable conformance vectors, and a reference validator.

The open protocol for persistent and verifiable governance of autonomous AI agents. Four-layer substrate: Persistent Context, Dynamic Governance, Deterministic Admissibility, Auditable Ledger. 0.07ms enforcement. Zero LLM inference.

KYA (Know Your Agents) — Open-source trust, governance, and evidentiary assurance infrastructure for autonomous systems. Built on KYP (Know Your Principal), a unified trust model for human users, AI agents, service accounts, and machine identities.

Sovereign Framework for Constitutional AI Governance — Five-layer architecture (L0-L4), three-cloud runtime, physics-certified contraction, multi-model consensus gate

Independent AI Agent Audit Framework - 包青天独立审计框架：穿透式风控体系，监督AI Agent的Skill调用合规性、防AI幻觉、系统性漏洞发现

The governance layer LLM proxies promised as a TypeScript library you can read. Every call runs through one governed pipeline: rate limit → spend caps (per-user + global circuit breaker) → cache → schema-validated provider failover → usage log → LLM-judge No sidecar proxy to deploy.

An AI Engineering Operating System for Claude Code — governed agents, repeatable skills, explicit permission gates, and a self-measuring telemetry layer. Windows-first. v1.

PyPI-distributable LLM control plane: gateway choke point, cost attribution, OTel instrumentation, and offline reporting as an inspectable engineering artifact.

Every AI agent action proven before it executes. Prolog + Z3 dual-gate, sub-millisecond verdicts, tamper-proof ledger. Apache 2.0.

Runtime approval gates for AI agent tool calls. Intercept payments and emails before execution

ZeroShield AI Mesh Firewall is a centralized AI security gateway for governing LLM and RAG traffic with inline prompt injection defense, vector database isolation, multi model routing control, and compliance grade observability.

ASHE — capability-broker protocol for AI agents. Open spec by Phor for AI agent capability mediation, multi-agent orchestration, and large language model governance. Apache 2.0 with patent grant.

Governance receipts for AI agents — MCP-native, fail-closed, LNN interpretability. Every action → compliance score → signed receipt → verifiable forever. EU AI Act Art. 12, Colorado AI Act (SB 24-205), NIST AI RMF.

Enterprise-grade governance and policy enforcement for agentic AI systems.

ICL (Imperative Context Language) is an open specification for governed AI agent execution in enterprise systems. It defines the event format for identity propagation, approval gating, rollback, and immutable audit — plus the cognitive trace that turns every deployment into labelled training data.

Bounded public demo of Aurora-Lens: stale, expired, or revoked evidence cannot carry decision authority.