QClaw Security Gateway: How It Protects Your Computer

Introduction

When an AI agent can read your files, send emails, and execute code on your computer, security isn't optional—it's essential. QClaw's AI Security Gateway (Claw Gateway) provides comprehensive protection, giving you complete visibility and control over every action.

This guide explains how the Security Gateway works and how to use it effectively.

Learn about QClaw security: https://qclawsg.qq.com

Understanding the Security Gateway

What is the Security Gateway?

The Security Gateway is QClaw's built-in security system that acts as a vigilant monitor for all AI operations. Think of it as:

• Security Camera: Records everything
• Permission Guard: Checks before executing
• Risk Assessor: Evaluates each action
• Alert System: Warns of suspicious activity
• Control Panel: Lets you set boundaries

Why It Matters

┌─────────────────────────────────────────────────────────────┐
│              Without Security Gateway                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│   User Request ──► AI Agent ──► Actions                     │
│                           │                                 │
│                           ▼                                 │
│                    No visibility                             │
│                    No control                               │
│                    Blind trust                              │
│                                                             │
└─────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────┐
│              With Security Gateway                           │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│   User Request ──► Security Gateway ──► AI Agent ──► Actions│
│                           │                                 │
│                           ▼                                 │
│                    ✓ Visibility                            │
│                    ✓ Control                               │
│                    ✓ Risk assessment                       │
│                    ✓ Full audit trail                      │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Core Components

1. Operation Monitor

Every action QClaw takes is logged:

2. Risk Assessment Engine

Actions are evaluated before execution:

# Risk scoring factors
class RiskAssessment:
    factors = {
        "data_sensitivity": "How sensitive is the data involved?",
        "action_reversibility": "Can this be undone?",
        "external_impact": "Does this affect outside systems?",
        "user_pattern": "Is this typical for this user?",
        "chain_analysis": "Does this fit the overall task?"
    }
    
    # Risk levels
    LOW = 0-30       # Auto-execute
    MEDIUM = 31-60   # Log and proceed
    HIGH = 61-80     # Require approval
    CRITICAL = 81-100 # Block by default

3. Permission Controller

Granular control over what QClaw can access:

4. Alert System

Notifications for important events:

Operation Types and Risk Levels

Low Risk (Auto-Approved)

These operations proceed automatically:

• Reading non-sensitive documents
• Opening specified applications
• Searching for information online
• Generating summaries
• Formatting text

Medium Risk (Logged)

These operations proceed but are logged:

• Creating new files
• Sending drafts for review
• Accessing Downloads folder
• Making non-sensitive changes
• Running approved scripts

High Risk (Approval Required)

These require your confirmation:

┌─────────────────────────────────────────────────────────────┐
│                    ⚠️ Approval Required                     │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  Action: Delete 15 files from Documents folder              │
│                                                             │
│  Files to be deleted:                                       │
│  • old_reports_2024.pdf                                     │
│  • draft_proposal_v1.docx                                   │
│  • meeting_notes_backup.txt                                 │
│  ... and 12 more                                            │
│                                                             │
│  Risk Assessment: High (67/100)                            │
│  Reason: Multiple file deletion                            │
│                                                             │
│  [Approve] [Deny] [Review Files]                           │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Critical Risk (Blocked)

These are blocked by default:

• Accessing system settings
• Installing applications
• Sending emails (without approval)
• Accessing sensitive folders
• Making network changes

Configuration Guide

Basic Security Settings

Access Settings: Settings → Security → Basic

Advanced Security Settings

For Power Users: Settings → Security → Advanced

security_config:
  # Permission Profiles
  profiles:
    strict:
      file_read: "Documents only"
      file_write: "Disabled"
      network: "Read-only"
      email: "Draft only"
      
    balanced:
      file_read: "Documents, Downloads"
      file_write: "Documents"
      network: "Allowed"
      email: "Draft + approve for send"
      
    permissive:
      file_read: "All folders"
      file_write: "Allowed"
      network: "Full"
      email: "Send with confirmation"

Folder Permissions

Granular Control:

You: "Allow QClaw to access my work projects folder"

QClaw: "Permission updated!

Allowed folders:
✓ Documents
✓ Downloads  
✓ ~/Projects/Work

Blocked folders:
✗ ~/Documents/Sensitive
✗ ~/.ssh
✗ ~/.config/credentials

Real-Time Dashboard

Star Office Security View

Access via the Star Office dashboard:

┌─────────────────────────────────────────────────────────────┐
│  QClaw Security Dashboard                                    │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  TODAY'S ACTIVITY                          Risk Distribution│
│  ┌────────────────────────────────┐     ┌───────────────┐  │
│  │ 47 operations completed         │     │ ██ 65% Low    │  │
│  │ 3 approvals requested          │     │ █ 25% Medium  │  │
│  │ 0 denied                       │     │ █ 8% High     │  │
│  │ 2 warnings flagged             │     │ █ 2% Blocked  │  │
│  └────────────────────────────────┘     └───────────────┘  │
│                                                             │
│  Recent Operations (Last 24h)                              │
│  ┌─────────────────────────────────────────────────────┐   │
│  │ 10:45 AM  ✓ Read   file.txt          Low     auto   │   │
│  │ 10:52 AM  ✓ Create summary.md        Medium   auto  │   │
│  │ 11:15 AM  ⚠️ Delete files (3)        High    pending│   │
│  │ 11:20 AM  ✓ Search web               Low     auto   │   │
│  └─────────────────────────────────────────────────────┘   │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Security Scenarios

Scenario 1: Unexpected File Access

Situation: QClaw tries to access a folder outside its permissions

Alert: Unauthorized Access Attempt

QClaw attempted to access: ~/Documents/Sensitive/HR/
Permission: Denied

This request was blocked. If this was unexpected:
• QClaw may have misinterpreted a request
• Check the original request for clarification

[View Details] [Allow This Once] [Update Permissions]

Scenario 2: Suspicious Pattern

Situation: Multiple file deletions in short succession

Warning: Elevated Activity Pattern

QClaw has performed 10 file operations in the last 5 minutes:
• 7 file reads
• 2 file creations
• 1 file deletion

This pattern differs from your normal usage.

Recommended Actions:
• Review recent operations
• Verify the current task context
• Continue monitoring

[Review Operations] [Pause QClaw] [All Normal]

Scenario 3: Email Without Permission

Situation: QClaw attempts to send an email

Critical: Email Send Blocked

QClaw attempted to: Send email via Gmail
Status: BLOCKED (email:send not permitted)

To allow email sending:
1. Settings → Security → Email
2. Enable "Allow email sending"
3. Set to "Approve each" or "Trusted senders only"

[Allow This Once] [Update Settings] [Dismiss]

Best Practices

For Individual Users

1. Start Restrictive

Initial Setup:
• file_read: Documents only
• file_write: Disabled
• email: Draft only
• network: Allowed

Gradually expand as you build confidence.

2. Review Regular Reports

Weekly Check:
✓ Review operation log
✓ Check for anomalies
✓ Verify all actions were intended
✓ Adjust permissions if needed

3. Enable Smart Alerts

Recommended Alerts:
• Any file deletion
• Any email action
• Network downloads
• Application installations
• System changes

For Business Users

1. Establish Policies

corporate_policy:
  default_profile: "strict"
  
  allowed_operations:
    file_read: ["Documents", "Shared"]
    file_write: ["Documents"]
    email_draft: true
    email_send: "manager_approval"
    
  blocked_operations:
    system_settings: true
    app_install: true
    sensitive_folders: ["HR", "Finance", "Legal"]

2. Regular Audits

Monthly Security Review:
• Operation statistics
• Risk incidents
• Permission usage
• Policy compliance
• User feedback

Transparency Features

Complete Audit Trail

Every operation is logged:

{
  "timestamp": "2026-04-29T14:32:15Z",
  "operation_id": "op_8f7a6b5c",
  "user_id": "user_abc123",
  "action": "file_delete",
  "target": {
    "path": "/Users/you/Documents/old_file.txt",
    "type": "text"
  },
  "risk_assessment": {
    "score": 85,
    "level": "high",
    "factors": ["multiple_files", "outside_work_folder"]
  },
  "status": "pending_approval",
  "context": {
    "task": "Clean up old files",
    "user_request": "organize my documents"
  }
}

Operation Replay

Review exactly what happened:

You: "Show me what QClaw did this morning"

QClaw: "Here's your morning activity:

9:00 AM - Task: Morning email summary
  ✓ Read inbox (50 emails)
  ✓ Starred 5 important
  ✓ Created summary document

10:30 AM - Task: Organize downloads
  ✓ Scanned Downloads folder
  ✓ Categorized 23 files
  ✓ Moved files to appropriate folders

12:00 PM - Task: Draft meeting response
  ✓ Read meeting invitation
  ✓ Created draft response
  ✓ Saved to draft folder

All operations completed successfully!"

Privacy Integration

Data Minimization

Third-Party Transparency

QClaw only shares:

┌─────────────────────────────────────────────────────────────┐
│              Data Shared with Third Parties                 │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ✓ Non-sensitive usage metrics (operation counts)           │
│  ✓ Error reports (no file content)                        │
│  ✓ Model interaction (input/output, not files)              │
│                                                             │
│  ✗ File contents - NEVER                                   │
│  ✗ Personal messages - NEVER                              │
│  ✗ Credentials - NEVER                                    │
│  ✗ Proprietary information - NEVER                         │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Responding to Alerts

Step-by-Step Response

1. Stay Calm

Security Gateway catches issues early—that's what it's designed for!

2. Assess the Context

• What task were you working on?
• Does this action make sense for that task?

3. Review the Details

• Check which files/apps are involved
• Verify the operation scope

4. Take Action

• Approve if legitimate
• Deny and pause if suspicious
• Report if concerning

5. Learn and Adjust

• Update permissions if needed
• Refine future prompts

Frequently Asked Questions

Q: Does Security Gateway slow down QClaw?

A: Minimal impact. Risk assessment adds <10ms to most operations.

Q: Can I disable Security Gateway?

A: No—it's a core component. But you can set it to auto-approve low-risk operations.

Q: What happens if QClaw does something I didn't approve?

A: You can review the operation log, revoke recent actions, and adjust permissions.

Q: Is my data safe with QClaw?

A: Yes. Files are processed locally, operation logs are encrypted, and Security Gateway monitors everything.

Q: Can I set different permissions for different tasks?

A: Current version uses session-wide permissions. Per-task permissions are planned.

Security Checklist

Initial Setup

• [ ] Review default security settings
• [ ] Configure folder permissions
• [ ] Enable notifications
• [ ] Test approval workflow
• [ ] Set up alert preferences

Ongoing Maintenance

• [ ] Weekly: Review operation logs
• [ ] Monthly: Check permission usage
• [ ] Quarterly: Update security policies
• [ ] As needed: Adjust for new use cases

Get Started Safely

QClaw's Security Gateway lets you harness AI power with complete peace of mind.

Download QClaw: https://qclawsg.qq.com

Related Articles

• QClaw Safety Guide: Best Practices
• QClaw Security Update
• QClaw vs Traditional AI Assistants

Your security is our priority: https://qclawsg.qq.com

Powerful AI with complete transparency and control.