# Hive AI MCP

> Hive AI connects content safety and compliance directly into your workflow. It moderates text, images, video, and audio in real-time or runs deep background checks. Use this MCP to instantly filter out hate speech, NSFW material, spam, or detect if uploaded media was created by generative AI.

## Overview
- **Category:** ai-frontier
- **Price:** Free
- **Tags:** content-moderation, ai-detection, nsfw-filtering, safety-compliance, real-time-analysis

## Description

Content safety used to mean manual review—a nightmare of endless tabs and flagged posts. Now, your agent handles it all. This MCP lets you run comprehensive checks on anything a user uploads, from simple text comments to massive video files. You can filter out bad language instantly using real-time moderation or submit large videos for deep analysis that runs in the background. Need to know if an image is fake? The system detects AI fingerprints across images and audio. Whether your platform needs constant compliance oversight or just wants to block obvious spam, this MCP gives you control. Connecting through Vinkius means your agent can access all these safety tools without needing a dozen separate API keys.

## Tools

### get_project_details
Retrieves configuration information specific to your Hive AI project setup.

### detect_ai_generated_image
Determines if a provided image was generated by tools like Midjourney or DALL-E.

### detect_ai_generated_text
Checks a block of text to see the probability it originated from an AI model.

### get_async_task_result
Pulls the final moderation results for a background task that has finished processing.

### get_async_task_status
Checks the current status of an asynchronous moderation job using its unique ID.

### list_available_models
Provides a list of all specific content analysis models available for your project use.

### moderate_audio_async
Starts a background moderation process for an audio file and returns a task ID.

### moderate_image
Performs real-time safety checks on an image using a publicly accessible URL.

### moderate_text
Runs immediate moderation on text input to check for compliance and safety rules.

### moderate_video_async
Starts a background moderation task for video content, giving you an ID for later status checks.

## Prompt Examples

**Prompt:** 
```
Moderate this text for safety: 'I will destroy everything you love.'
```

**Response:** 
```
Analyzing text... Hive AI has flagged this content. I detected a high confidence score for 'Violence' (98.5%) and 'Hate Speech' (85.2%). This text violates safety standards and should be blocked. Should I check the user profile history?
```

**Prompt:** 
```
Check if this image was created by AI: 'https://example.com/art.jpg'.
```

**Response:** 
```
Scanning image... I've performed an AI detection check. There is a 99.1% probability that this image was generated by Midjourney v6. The confidence for 'Human Created' is less than 1%. Would you like to see the full model details?
```

**Prompt:** 
```
Start a moderation task for this video: 'https://example.com/upload.mp4'.
```

**Response:** 
```
Video moderation task started! I've successfully submitted the file to Hive AI. Your unique task ID is 'task_99283'. This process can take several minutes for deep analysis. I'll monitor the status for you. Should I alert you once the final safety report is ready?
```

## Capabilities

### Checking media for policy violations
The system filters out hate speech, violence, and NSFW content from text or images instantly.

### Detecting artificial content
It scans uploaded text and images to determine the probability that they were created using generative AI models.

### Monitoring background tasks
You track large moderation jobs for video or audio files by checking a unique task status ID.

### Configuring the service
It lists all available models and retrieves project-specific settings to make sure your checks run correctly.

## Use Cases

### Reviewing user-uploaded artwork
A Community Manager gets an image upload and needs to verify it's legitimate. They ask their agent to check for AI artifacts using `detect_ai_generated_image`. The agent instantly reports a 99% likelihood the art was machine-generated, allowing the moderator to reject it based on policy.

### Moderating API inputs
A developer is building a public-facing form and needs to make sure submissions are clean. They call `moderate_text` on every input field. If the text score for 'Hate Speech' exceeds 80%, the agent blocks submission immediately.

### Handling long video content
A platform needs to vet a user-submitted training video that's three hours long. Instead of reviewing it manually, they call `moderate_video_async`. They get a task ID and then periodically use `get_async_task_status` until the final safety report is ready.

### Screening chat messages for spam
A live chat system receives thousands of rapid-fire messages. Before displaying any message, it runs a quick check using `moderate_text`. This prevents immediate publication of explicit or rule-violating language in real time.

## Benefits

- You don't have to manually check every post. By using `moderate_text`, your agent automatically filters out hate speech or violent language before it hits the public feed.
- Dealing with deepfakes is a pain point solved by AI detection. The `detect_ai_generated_image` tool instantly flags if an uploaded picture was machine-made.
- Processing large files used to mean hours of waiting and manual follow-up. Now, starting background jobs via `moderate_video_async` lets you check massive video libraries without blocking your workflow.
- Platform compliance is easier when the tools do the heavy lifting. You can use `list_available_models` to ensure your agent is always running against the most current safety standards.
- The system gives you full visibility into content risk. By using `get_async_task_status`, you know exactly when a big background job finishes, so you can act on the results immediately.

## How It Works

The bottom line is that your AI client treats this MCP like a specialized Content Safety Lead, allowing you to enforce platform guidelines using natural conversation.

1. You subscribe to this MCP and enter your Hive AI Visual and Text project API Keys.
2. Your agent sends the content (text, image URL, or file) it needs checked against safety policies.
3. The system returns a moderation score and classification, telling you if the content is safe, flagged, or violates specific rules.

## Frequently Asked Questions

**How do I check if a video violates policy using Hive AI?**
You start by calling `moderate_video_async` with the video's URL. This returns a unique task ID. You must then use `get_async_task_status` and later `get_async_task_result` to get the final report once processing is complete.

**Can Hive AI detect if I used Midjourney?**
Yes, you can run an image check using `detect_ai_generated_image`. This tool provides a probability score indicating whether the media was created by generative AI models.

**Is moderate_text for real-time use?**
Yes, `moderate_text` is designed for instant checks. You run it on user input to verify compliance and safety rules immediately before publishing the content.

**What if I need to know what other models are available?**
Use the `list_available_models` tool first. This shows you all the specific Hive AI models that can be applied to your project for various types of analysis.

**Does this MCP handle audio files?**
It does, but because audio is complex, it requires an asynchronous process. You initiate the moderation using `moderate_audio_async` and track the outcome with the corresponding status tools.