MCP Server Integration

5 min read

Overview

Cacuda provides an MCP (Model Context Protocol) server that AI assistants like Cursor, Claude Desktop, and others can use to moderate content directly from within their workflows.

This enables AI-powered agents to check content for spam, NSFW material, and other policy violations as part of their reasoning process.

Installation

Add the Cacuda MCP server to your AI assistant's configuration. Here's the configuration block for common tools:

Cursor / Claude Desktop

Add to your MCP settings file:

{
  "mcpServers": {
    "cacuda": {
      "command": "npx",
      "args": ["@cacuda/mcp-server"],
      "env": {
        "CACUDA_API_KEY": "your-api-key",
        "CACUDA_API_URL": "https://api.cacuda.com"
      }
    }
  }
}

npm install

Or install globally:

npm install -g @cacuda/mcp-server

Configuration

The MCP server requires two environment variables:

CACUDA_API_KEY — your Cacuda API key (generate at /account)
CACUDA_API_URL — the API base URL (defaults to https://api.cacuda.com)

Available Tools

The MCP server exposes 5 tools to AI assistants:

moderate_content

Run content through a full automod pipeline and return the moderation result.

check_text

Analyze text against specific categories (spam, toxicity, NSFW) without running a full pipeline.

check_image

Analyze an image URL for NSFW content, violence, or other visual policy violations.

list_automods

List all automods in your account with their status and configuration.

get_execution

Retrieve details and step-by-step results from a previous moderation execution.

Example Usage

An AI assistant can call the tools like this:

Check text content

// Tool call
{
  "tool": "check_text",
  "arguments": {
    "text": "Buy cheap products now! Click here!!!",
    "categories": ["spam", "toxicity"]
  }
}

// Response
{
  "results": {
    "spam": { "score": 0.92, "label": "spam" },
    "toxicity": { "score": 0.05, "label": "not_toxic" }
  }
}

Run a full automod pipeline

// Tool call
{
  "tool": "moderate_content",
  "arguments": {
    "automodId": "am_abc123",
    "content": "Check this user-submitted comment"
  }
}

// Response
{
  "executionId": "exec_xyz789",
  "status": "completed",
  "action": "approve",
  "creditsUsed": 1
}

Using Automod as MCP Trigger

You can also configure an automod with an MCP trigger node type. This allows AI agents to invoke your full moderation pipeline through the moderate_content tool, running content through all your configured detection and action nodes.

Tip: This is ideal for building AI-powered content workflows where an agent gathers content, moderates it through Cacuda, and takes action based on the result.

Next: Word Lists