NexusClawHQ/nexusclaw-agent-governance

README 摘要

agent-governance English 中文 The deny-by-default governance kernel for AI agents. Permissions · L0–L4 guardrails · human approvals · immutable audit chain — as an Apache-2.0, framework-neutral library. Built by NexusClaw — the AI-native CRM with governed digital employees — from the kernel that runs its production executions. Quick start · Packages · Adapters · Reference implementation · 中文 What this is Every agent execution starts with nothing allowed : no tool, no object, no operation. Access is granted only through an explicit, auditable policy — and every grant, denial, approval and execution step is written to an immutable audit chain . That is the problem this library solves — and it solves it without owning your agent. Keep your framework (LangGraph, CrewAI, n8n, Dify, or a plain script), route your tool calls through the gate API, and the kernel handles permissions, risk evaluation, human approval pauses and audit records. This repository is the open-source (Apache-2.0) governance kernel that runs NexusClaw's digital employees in production, shipped in three forms: 1. Library — nine framework-neutral npm packages under governance/ (zero-dependency core packages), plus framew…

分类