docs/MCP.md

# MCP (Model Context Protocol) Integration

This Space is accessible via the Model Context Protocol, allowing integration with MCP clients like Cursor, Claude Desktop, and other compatible tools.

## Configuration

### For Claude Desktop

Add the following to your Claude Desktop MCP configuration file:

**macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows:** `%APPDATA%\Claude\claude_desktop_config.json`

```json
{
  "mcpServers": {
    "flux-texture-lora": {
      "url": "https://huggingface.co/spaces/gokaygokay/Flux-Seamless-Texture-LoRA/gradio_api/mcp/",
      "type": "http"
    }
  }
}
```

### For Cursor

Add to your Cursor MCP configuration:

```json
{
  "mcpServers": {
    "flux-texture-lora": {
      "url": "https://huggingface.co/spaces/gokaygokay/Flux-Seamless-Texture-LoRA/gradio_api/mcp/",
      "type": "http"
    }
  }
}
```

## Available Tools

### `generate_texture`

Generate a texture image with specified parameters.

**Parameters:**
- `prompt` (str): Text prompt for generation
- `negative_prompt` (str, optional): Negative prompt
- `guidance_scale` (float, optional): Guidance scale (default: 7.5)
- `num_inference_steps` (int, optional): Number of steps (default: 50)
- `seed` (int, optional): Random seed
- `width` (int, optional): Image width (default: 1024)
- `height` (int, optional): Image height (default: 1024)

**Returns:**
- `dict`: Contains `success`, `image_path`, `seed`, and `message`

### `generate_batch_textures`

Generate multiple textures in batch.

**Parameters:**
- `prompts` (List[str]): List of prompts
- `guidance_scale` (float, optional): Guidance scale
- `num_inference_steps` (int, optional): Number of steps

**Returns:**
- `dict`: Contains `success`, `total`, and `results` list

### `get_presets`

Get list of available presets.

**Returns:**
- `dict`: Contains `presets` list and `details` dictionary

### `get_history`

Get generation history.

**Parameters:**
- `limit` (int, optional): Maximum entries to return (default: 10)

**Returns:**
- `dict`: Contains `total` and `entries` list

### `download_image`

Get information about an image for download.

**Parameters:**
- `image_path` (str): Path to the image

**Returns:**
- `dict`: Contains `success`, `image_path`, and `metadata`

## Available Resources

### `presets://`

Access preset definitions as JSON.

**Example:**
```
presets://wood-texture
```

### `history://`

Access generation history as JSON.

**Example:**
```
history://recent
```

### `examples://`

Access example images information.

**Example:**
```
examples://list
```

## Usage Examples

### Using with Claude Desktop

1. Configure MCP in Claude Desktop settings
2. Restart Claude Desktop
3. Ask Claude to generate a texture:
   ```
   Generate a wood texture using the flux-texture-lora tool
   ```

### Using with Cursor

1. Configure MCP in Cursor settings
2. Use the MCP tools in your code:
   ```python
   # Cursor can now access the MCP tools
   # to generate textures programmatically
   ```

## Troubleshooting

### Connection Issues

- Verify the Space URL is correct
- Check that the Space is running
- Ensure MCP is enabled in the Space configuration

### Tool Errors

- Check parameter types and ranges
- Verify prompts are not empty
- Ensure image dimensions are multiples of 8

## API Endpoint

The MCP endpoint is available at:
```
/gradio_api/mcp/
```

Full URL:
```
https://huggingface.co/spaces/gokaygokay/Flux-Seamless-Texture-LoRA/gradio_api/mcp/
```