You have a folder of product photos that need to be print-ready. Or AI-generated images that came out a little soft. Or a client asking for “higher resolution” on everything you’ve sent them this week.
You could open each image in Photoshop, export, upload to an upscaling service, wait, download, rename, repeat. Or you could just tell Claude:
“Upscale all the images in my /products folder to 4x.”
That’s it. Claude reads the folder, sends each image through PixelPanda’s AI upscaler, and saves the results. No tabs, no uploads, no waiting around watching progress bars.
This guide walks you through the setup — about 3 minutes — and then shows you what’s possible.
What You’ll Need
- Claude Desktop or Claude Code (any MCP-compatible client works)
- Python 3.10+ installed on your machine
- A PixelPanda account with credits (the $5 Try It pack works)
Step 1: Install the MCP Server
Open your terminal and run:
pip install pixelpanda-mcp
That’s the entire installation. The package includes 33 AI image tools — upscaling, enhancement, background removal, product photography, and more.
Step 2: Get Your API Token
- Go to pixelpanda.ai and log in
- Click your profile icon in the top right
- Click API Token
- Click Generate Token
- Copy the token — it starts with
pk_live_
Keep this token private. It’s tied to your account and credits.
Step 3: Connect to Claude
Claude Desktop
Open your Claude Desktop config file:
- Mac:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add the PixelPanda server:
{
"mcpServers": {
"pixelpanda": {
"command": "pixelpanda-mcp",
"env": {
"PIXELPANDA_API_TOKEN": "pk_live_your_token_here"
}
}
}
}
Restart Claude Desktop.
Claude Code
Add to ~/.claude/claude_code_config.json:
{
"mcpServers": {
"pixelpanda": {
"command": "pixelpanda-mcp",
"env": {
"PIXELPANDA_API_TOKEN": "pk_live_your_token_here"
}
}
}
}
Restart Claude Code. You should see PixelPanda’s tools in the available tools list.
Upscaling Images
Basic Upscale
Just tell Claude what you want:
“Upscale this image to 4x: https://example.com/my-photo.jpg”
Claude calls PixelPanda’s AI upscaler and returns the result URL. The upscaled image is stored permanently — you can download it anytime.
Upscale a Local File
“Upscale ~/Desktop/product-shot.png to 4x”
Claude reads the file from your machine, sends it through the upscaler, and gives you the result.
Print-Ready 8x
For large format printing (posters, banners, trade show displays):
“I need this image print-ready at 300 DPI for a 24×36 poster. Upscale it as much as needed.”
Claude will calculate the required resolution and choose the right scale factor.
Batch Upscale a Folder
“Upscale every PNG in my ~/products/photos folder to 4x”
Claude iterates through the folder and upscales each image. You get back a list of result URLs.
Scale Options and Pricing
| Scale | Credits | Speed | Best For |
|---|---|---|---|
| 2x | 1 | ~5-8 seconds | Quick size bump |
| 4x | 3 | ~30-60 seconds | E-commerce, social media |
| 8x | 5 | ~2-3 minutes | Print, large format |
The 2x option uses Real-ESRGAN (fast). The 4x and 8x options use Clarity Upscaler, which handles any image size through intelligent tiling — no GPU memory limits.
Quality Modes
You can also specify a quality mode:
“Upscale this to 2x using high quality mode”
- fast — Real-ESRGAN. Quick and clean. Best for images that are already decent quality.
- balanced — Clarity Upscaler with optimized settings. Good tradeoff between speed and quality.
- high — Clarity Upscaler at full quality. Maximum detail preservation. Worth it for hero images.
Enhancing Images
Enhancement is different from upscaling — it sharpens, denoises, restores faces, and fixes artifacts. PixelPanda runs 4 different AI enhancement methods in parallel and gives you all the results so you can pick the best one.
Basic Enhance
“Enhance this image: https://example.com/blurry-photo.jpg”
Claude returns multiple variants:
- Clean Upscale — Faithful sharpening without changing the image character
- Face Enhance — Extra detail on faces (great for portraits and headshots)
- AI Restore — Fixes compression artifacts, scratches, and damage
- Full Restore — The most aggressive option — restores and then upscales with face enhancement
Each enhance costs 1 credit and you get all 4 variants.
Enhance AI-Generated Images
AI-generated product photos sometimes come out a little soft or have minor artifacts. Enhancement fixes that:
“These AI product photos look slightly soft. Can you enhance them?”
Fix Old or Damaged Photos
“This photo is from 2005 and looks pretty rough. Can you restore it?”
The AI Restore and Full Restore variants are specifically designed for this — they reconstruct missing detail and fix compression artifacts.
Combining Upscale + Enhance
The real power is chaining operations:
“Enhance this photo first, then upscale the best result to 4x for printing”
Claude runs the enhancement, picks the best variant, and feeds it into the upscaler. Two operations, one sentence.
Other Things You Can Do
Once you have the MCP connected, you’re not limited to upscaling and enhancement. The same connection gives you:
- Remove backgrounds — “Remove the background from this product photo”
- Generate product photos — “Create 6 marketing photos of my sunglasses with the lifestyle avatar”
- Virtual try-on — “Show this dress on my fashion avatar”
- Scene generation — “Put this coffee maker in a modern kitchen setting”
- Check your credits — “How many credits do I have left?”
All through natural conversation with Claude.
Troubleshooting
“Tool not found” or Claude doesn’t see PixelPanda tools
– Make sure you restarted Claude after editing the config
– Check that pixelpanda-mcp is installed: pip show pixelpanda-mcp
– Verify the command path: which pixelpanda-mcp
“Invalid or missing API token”
– Double-check your token starts with pk_live_
– Make sure it’s in the env section of your MCP config, not args
– Regenerate the token at pixelpanda.ai if it’s not working
“Insufficient credits”
– Check your balance: ask Claude “How many PixelPanda credits do I have?”
– Top up at pixelpanda.ai/pricing
Upscale is slow
– 4x and 8x use Clarity Upscaler which takes 30-180 seconds depending on image size
– 2x uses Real-ESRGAN and finishes in 5-8 seconds
– Large images (4K+) take longer due to tiling
What’s Next
You’ve got a direct line from Claude to professional image processing. No context-switching, no manual uploads, no browser tabs. Just say what you need and it happens.
Start with one image. Upscale it. Enhance it. Then try a batch. Once you see the workflow, you won’t go back to doing it manually.
Get started: pixelpanda.ai/pricing