AutoGuard AI

A comprehensive AI agent monitoring and safety system that tracks agent behaviors, assesses risks in real-time, and provides alerts when agents cross safety thresholds.

Features

• Agent Registration: Register AI agents with unique API keys for monitoring
• Behavior Logging: Track and log agent actions with risk scores
• Risk Assessment: Real-time risk evaluation with multiple factors
• Alert System: Automatic alerts when risk levels reach high/critical thresholds
• Real-time Dashboard: Live monitoring with WebSocket updates
• Data Visualization: Interactive charts for risk trends and behavior analysis
• Settings Management: Configurable risk thresholds and notification preferences

Architecture

autoguard-ai/
├── apps/
│   ├── api/          # Express.js backend API
│   └── web/          # React frontend dashboard
└── packages/
    └── shared/       # Shared TypeScript types and utilities

Tech Stack

• Backend: Node.js, Express, TypeScript, WebSocket (ws)
• Frontend: React, TypeScript, TailwindCSS, Recharts
• Build: pnpm workspaces, Vite

Getting Started

Prerequisites

• Node.js 18+
• pnpm 8+

Installation

# Clone the repository
git clone <repository-url>
cd autoguard-ai

# Install dependencies
pnpm install

# Build shared package
pnpm --filter @autoguard-ai/shared build

Development

# Start API server (from root)
pnpm --filter @autoguard-ai/api dev

# Start web dev server (in another terminal)
pnpm --filter @autoguard-ai/web dev

The API server runs on http://localhost:3000 and WebSocket on ws://localhost:3001.

The web app runs on http://localhost:5173.

Production Build

# Build all packages
pnpm build

# Or build individually
pnpm --filter @autoguard-ai/api build
pnpm --filter @autoguard-ai/web build

API Documentation

Authentication

Most endpoints require an API key in the X-API-Key header. Get your API key when registering an agent.

Endpoints

Agents

Method	Endpoint	Description
GET	/api/agents	List all agents
GET	/api/agents/:id	Get agent by ID
POST	/api/agents	Create new agent
PUT	/api/agents/:id	Update agent
DELETE	/api/agents/:id	Delete agent

Behavior Logs

Method	Endpoint	Description
POST	/api/logs	Submit behavior log (auth required)
GET	/api/agents/:id/logs	Get agent's behavior logs

Risk Assessment

Method	Endpoint	Description
GET	/api/agents/:id/risk	Get agent's risk assessment

Alerts

Method	Endpoint	Description
GET	/api/alerts	List all alerts
GET	/api/alerts/recent	Get recent alerts
GET	/api/alerts/:id	Get alert by ID
POST	/api/alerts	Create alert
POST	/api/alerts/:id/acknowledge	Acknowledge alert
POST	/api/alerts/:id/resolve	Resolve alert

Settings

Method	Endpoint	Description
GET	/api/settings	Get all settings
PUT	/api/settings	Update settings
GET	/api/settings/thresholds	Get risk thresholds
PUT	/api/settings/thresholds	Update thresholds

Example: Register an Agent

curl -X POST http://localhost:3000/api/agents \
  -H "Content-Type: application/json" \
  -d '{"name": "My Agent", "description": "A helpful AI assistant"}'

Response:

{
  "success": true,
  "data": {
    "id": "1234567890-abc123",
    "name": "My Agent",
    "description": "A helpful AI assistant",
    "apiKey": "ag-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
    "status": "active",
    "riskLevel": "low",
    "createdAt": "2024-01-01T00:00:00.000Z",
    "updatedAt": "2024-01-01T00:00:00.000Z"
  }
}

Example: Submit a Behavior Log

curl -X POST http://localhost:3000/api/logs \
  -H "Content-Type: application/json" \
  -H "X-API-Key: ag-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -d '{
    "agentId": "agent-id",
    "action": "data_access",
    "description": "Accessed user profile data",
    "riskScore": 25
  }'

Risk Levels

Level	Score Range	Description
Low	0-24	Normal, safe behavior
Medium	25-49	Minor concerns, monitor closely
High	50-74	Significant risk, review recommended
Critical	75-100	Severe risk, immediate action required

Risk Factors

The system evaluates risk based on multiple factors:

1. Activity Frequency: Number of recent actions
2. High-Risk Actions: Count of actions with high risk scores
3. Behavior Consistency: Variance in behavior patterns
4. Risk Trend: Direction of risk level changes

WebSocket Events

Connect to ws://localhost:3001 for real-time updates:

const ws = new WebSocket('ws://localhost:3001');

ws.onmessage = (event) => {
  const message = JSON.parse(event.data);
  console.log('Event:', message.type, message.payload);
};

Event Types

• agent:created - New agent registered
• agent:updated - Agent details changed
• agent:deleted - Agent removed
• behavior:logged - New behavior log submitted
• risk:updated - Agent risk level changed
• alert:created - New alert triggered
• alert:acknowledged - Alert acknowledged
• alert:resolved - Alert resolved

Configuration

Environment Variables

# API Server
PORT=3000

# WebSocket Server (runs on port + 1)
WS_PORT=3001

Settings

Access the Settings page in the dashboard to configure:

• Risk Thresholds: Customize score boundaries for each risk level
• Alert Rules: Enable/disable automatic alert triggers
• Notifications: Configure email and webhook notifications

Project Structure

/apps/api

Express.js backend with:

• RESTful API routes
• WebSocket server for real-time updates
• In-memory storage (can be extended to use databases)
• Risk calculation algorithms

/apps/web

React frontend with:

• Dashboard with real-time charts
• Agent registration and management
• Detailed agent view with risk assessment
• Settings configuration
• Responsive design with TailwindCSS

/packages/shared

Shared TypeScript types and utilities:

• Type definitions for all data models
• API request/response types
• Risk calculation functions
• ID and API key generation

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

License

MIT