Clark Documentation (NOT FINISHED)

Introduction

Clark is built on a proprietary Neural Threat Detection model to identify and neutralize malicious accounts with zero-second response time. Unlike traditional bots that react after a member joins, Clark analyzes accounts before they can cause any harm, offering seamless, automated protection for your community.

Installation & Setup

1. Add Clark to your server

You can invite Clark using the following link. Please ensure you are logged into your Discord account before clicking:

https://discord.com/oauth2/authorize?client_id=1422636332454514779

2. Required Permissions

For Clark to function optimally, it requires certain permissions. Granting the Administrator permission is highly recommended for full protection against raids and to ensure it can enforce policies instantly. At a minimum, Clark requires:

• Kick Members (To remove low-risk, suspicious accounts)
• Ban Members (To neutralize confirmed threats and high-risk accounts)
• Manage Guild (For full configuration access)

Read more about configuring permissions on our Permissions page.

AI Join Detection

This is Clark's flagship feature. Every new account is instantly assessed based on hundreds of factors. If the AI decides that the user is suspicious, it will flag that user.

Tiers & Actions

You can set custom actions for different score ranges via the Web Dashboard:

• 0-50 (Low Risk): Typically no action taken.
• 51-80 (Medium Risk): Recommended action is Kick and logging the event.
• 81-100 (High Risk): Recommended action is an immediate Ban.

Customizing the Threshold (Not available yet)

Adjust the threshold using a simple command in your Discord server:

/flag setup set threshold (amount)

This sets the default action (e.g., ban) to occur if an account's risk score is 85 or higher.

Anti-Raid System

Our Anti-Raid system monitors join velocity, detecting bursts of new accounts that signal a coordinated attack. When a raid is detected, Clark automatically enters lockdown mode.

Lockdown Mode

In lockdown, new users are prevented from joining, and the detection threshold is temporarily increased to its maximum sensitivity until the threat passes. All raid events are logged on the Web Dashboard.