🛡️ MoltGuard

Security & observability platform for AI agents.

Complete visibility and control over your AI agents. See what they're thinking, approve risky actions, stop them when needed.

🌐 Live Demo

https://guard.moltnet.ai

✨ Features

🔍 Skill Scanner

Analyze skills for malicious patterns before installation:

• Remote code execution detection
• Data exfiltration patterns
• Credential access attempts
• Obfuscation detection
• Supply chain risks

🧠 Mind Graph

Visualize what your agent is thinking:

• Real-time thought logging
• Decision tree visualization
• Session timeline
• Reasoning trace

✋ Intent Gating

Require approval for risky actions:

• Pending action queue
• Approve/reject workflow
• Auto-timeout
• Webhook callbacks
• Telegram notifications

🎮 Remote Control

Control your agents in real-time:

• Pause / Resume
• Emergency Stop
• Command history
• Multi-agent management

📋 Audit Log

Complete action history:

• Search and filter
• Export (JSON/CSV)
• Cost tracking
• Token usage

🏷️ Badges

Show skill safety in your README:

![MoltGuard](https://guard.moltnet.ai/api/badge?url=YOUR_SKILL_URL)

📄 Pages

Page	Description
/	Landing page with quick scan
/scan	Single skill scanner
/batch	Batch scan (up to 10)
/compare	Side-by-side comparison
/dashboard	Activity log & stats
/mind-graph	Thought visualization
/control	Remote agent control
/architecture	How it works
/docs	API documentation
/skill.md	Clawdbot integration skill

🔌 API Endpoints

Skill Scanner

• POST /api/scan — Scan a skill
• POST /api/batch-scan — Scan multiple skills
• POST /api/compare — Compare two skills
• GET /api/badge?url= — Get SVG badge

Action Logging

• POST /api/actions — Log an action
• GET /api/pending — Get pending actions
• POST /api/actions/:id/approve — Approve
• POST /api/actions/:id/reject — Reject

Mind Graph

• POST /api/sessions — Start session
• POST /api/traces — Log thought
• GET /api/mind-graph/:sessionId — Get graph data

Remote Control

• POST /api/agents/register — Register agent
• POST /api/control/:agent — Send command
• GET /api/control/:agent/pending — Get commands

🚀 Quick Start

For AI Agents (Clawdbot/OpenClaw)

1. Register your agent:

curl -X POST https://guard.moltnet.ai/api/agents/register \
  -H "Content-Type: application/json" \
  -d '{"name": "my-agent", "description": "My AI assistant"}'

2. Log your thoughts:

curl -X POST https://guard.moltnet.ai/api/traces \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "my-agent",
    "type": "reasoning",
    "title": "Analyzing request",
    "content": "User wants me to..."
  }'

3. Request approval for risky actions:

curl -X POST https://guard.moltnet.ai/api/actions \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "my-agent",
    "type": "email.send",
    "description": "Send report",
    "risk": "high",
    "status": "pending"
  }'

4. Check for control commands:

curl https://guard.moltnet.ai/api/control/my-agent/pending

For Humans

1. Open https://guard.moltnet.ai
2. View your agent's activity in the Dashboard
3. See their thinking in Mind Graph
4. Control them from Remote Control

🔧 Self-Hosting

git clone https://github.com/moltnet/guard
cd moltguard
npm install
npm start

Environment Variables

Variable	Description	Default
PORT	Server port	3457
MOLTGUARD_API_KEY	API authentication	none
TELEGRAM_BOT_TOKEN	Telegram alerts	none
TELEGRAM_CHAT_ID	Telegram chat	none
PENDING_TIMEOUT_MS	Auto-reject timeout	300000

📊 Architecture

┌─────────────┐      ┌─────────────┐      ┌─────────────┐
│  AI Agent   │ ───▶ │  MoltGuard  │ ◀─── │    Human    │
│  (Clawdbot) │      │  (Security) │      │ (Dashboard) │
└─────────────┘      └─────────────┘      └─────────────┘
       │                    │                    │
       │  Logs thoughts     │  Stores data       │  Views activity
       │  Requests approval │  Sends alerts      │  Approves/rejects
       │  Checks commands   │  Enforces policy   │  Sends commands
       ▼                    ▼                    ▼
┌─────────────────────────────────────────────────────────┐
│                     SQLite Database                      │
│  actions | traces | sessions | agents | commands | scans │
└─────────────────────────────────────────────────────────┘

📄 License

MIT

---

Built with ⚡ by @rohansxd