Add text watermarks to images and PDFs programmatically via AI agents. Runs locally — files never leave your machine.
Install the MCP server and configure your AI assistant to use it.
npx valepara-mcp
Add to ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"valepara": {
"command": "npx",
"args": ["-y", "valepara-mcp"]
}
}
}claude mcp add valepara-mcp -- npx -y valepara-mcp
Add a repeating text watermark to an image file (JPG, PNG, WebP). Returns the watermarked file path and a preview image.
| Parameter | Type | Default | Description |
|---|---|---|---|
| imagePath | string | required | Absolute path to the source image |
| outputPath | string | auto | Output path. Defaults to {name}-watermarked.{ext} |
| text | string | "ONLY VALID FOR" | Watermark text to repeat |
| opacity | number | 0.4 | Opacity (0-1) |
| fontSize | number | 28 | Font size in pixels |
| rotation | number | -30 | Rotation angle (-90 to 90 degrees) |
| color | string | "#ff0000" | Hex color |
| spacing | number | 100 | Gap between repetitions in px |
| wavy | boolean | false | Apply wavy text effect |
| wavyAmplitude | number | 8 | Wave height in px |
| wavyFrequency | number | 0.05 | Wave oscillation frequency |
Use watermark-image on /Users/me/id-card.jpg with text "COPY - NOT VALID" opacity 0.5 color "#ff0000" fontSize 32
Add a repeating text watermark to all pages of a PDF file. Same parameters as watermark-image but with pdfPath instead of imagePath.
Use watermark-pdf on /Users/me/contract.pdf with text "DRAFT - CONFIDENTIAL" color "#888888" fontSize 48 opacity 0.3
Detect the document boundary in a photo using Canny edge detection. Returns 4 corner points (TL, TR, BR, BL) or null if no document is found.
| Parameter | Type | Default | Description |
|---|---|---|---|
| imagePath | string | required | Path to a photo containing a document |
{
"detected": true,
"corners": [
{ "x": 120, "y": 85 },
{ "x": 890, "y": 90 },
{ "x": 885, "y": 620 },
{ "x": 115, "y": 615 }
],
"imageWidth": 1024,
"imageHeight": 768
}Ask your AI assistant to watermark multiple files at once:
Watermark all JPG files in /Users/me/documents/ with text "PERSONAL COPY" opacity 0.3 color "#0000ff"
First detect the document area, then apply watermark only to that region:
1. detect-document on /Users/me/photo-of-passport.jpg 2. If detected, watermark-image with the bounds text "VALID FOR RENTAL ONLY" rotation -30
Watermark sensitive documents before sending:
Watermark /Users/me/tax-return.pdf with text "SHARED WITH: Acme Corp - March 2026" opacity 0.2 fontSize 16 color "#999999"
The MCP server runs entirely on your machine. No files are uploaded to any external service. Watermarking happens in-process using Sharp (images) and pdf-lib (PDFs). Your documents never leave your computer.