DEFCON ONE™

Human-authority-first AI execution control system

AI may propose. Humans must approve. Systems must obey.

Overview

DEFCON ONE™ is a universal authority gate for AI agents. It ensures that AI systems cannot execute protected actions without explicit human approval, while maintaining privacy and enabling self-hosted deployments.

Features

• ✅ Universal Plugin Architecture - Works with ANY AI agent
• ✅ MCP Protocol Support - Native Claude Code integration
• ✅ Python & TypeScript SDKs - Easy integration for developers
• ✅ Ed25519 Cryptographic Tokens - Single-use, time-limited
• ✅ Complete Audit Trail - Every action logged
• ✅ Privacy-First - Self-hostable, no corporate surveillance
• ✅ Two-Lane Model - Lane A (unrestricted) vs Lane B (protected)
• ✅ Docker Deployment - One-command setup
• ✅ Production Ready - Tested and operational

Quick Start

See DEFCON_INTEGRATION_GUIDE.md for complete instructions.

Documentation

• Integration Guide - Complete integration guide
• Plugin System - Plugin architecture overview
• Architecture - System architecture details

Status

• Backend: ✅ Operational (port 7100)
• Reactor Integration: ✅ Tested and working
• MCP Server: ✅ Ready for Claude Code
• Python SDK: ✅ Available
• TypeScript SDK: ✅ Available
• Docker Deployment: ✅ Available
• Documentation: ✅ Complete
• Test Results: ✅ 8/8 tests passing

License

MIT License

Support

• WOPR Systems: https://wopr.systems
• Forgejo Vault: http://vault.wopr.systems:3001

---

Built with the philosophy of human authority over AI execution.

Co-Authored-By: Claude Sonnet 4.5