How MoltGuard Works

Complete visibility and control over your AI agents.

🤖
AI Agent
Clawdbot / OpenClaw
↓ Actions & Thoughts
🛡️
MoltGuard
Security Layer
↓ Logs, Alerts, Approvals
👤
You
Dashboard & Control
🔍 Skill Protection
1

Scan Before Install

Before your agent installs any skill, MoltGuard analyzes it for malicious patterns, data exfiltration, and code injection.

2

Get Risk Score

Each skill gets a risk score (0-100) based on detected patterns. Safe skills score low; dangerous ones score high.

3

Make Informed Decisions

See exactly what risks exist before installing. Compare skills side-by-side.

✋ Intent Gating
1

Agent Requests Action

When your agent wants to do something risky (send email, delete file, make purchase), it sends the request to MoltGuard.

2

You Get Notified

MoltGuard alerts you via dashboard or Telegram. The agent waits for your decision.

3

Approve or Reject

You review the action and approve or reject it. The agent receives your decision and proceeds accordingly.

🧠 Mind Graph
1

Agent Logs Thoughts

Your agent sends its reasoning process to MoltGuard — what it's thinking, why it's making decisions.

2

Visualize Decision Tree

See a real-time graph of the agent's thought process, from initial prompt to final action.

3

Understand & Debug

When something goes wrong, trace back through the reasoning to understand why.

🎮 Remote Control
1

Pause / Resume

Instantly pause an agent if something looks wrong. Resume when you're satisfied.

2

Emergency Stop

Send a kill command to immediately stop all agent activity.

3

Set Boundaries

Configure what your agent can and cannot do. Set spending limits, blocked actions, approved domains.

Quick Integration

# Your agent sends actions to MoltGuard
curl -X POST http://82.112.226.62:3457/api/actions \
  -d '{"agent":"my-agent","type":"email.send","risk":"high","status":"pending"}'

# Your agent logs its thoughts
curl -X POST http://82.112.226.62:3457/api/traces \
  -d '{"agent":"my-agent","type":"reasoning","title":"Deciding next step","content":"..."}'

# Your agent checks for control commands
curl http://82.112.226.62:3457/api/control/my-agent/pending