Product photos stolen. Blog images reposted without credit. Client deliverables ripped and resold. If you publish images on the web, you've either dealt with unauthorized reuse or you will.
Watermarks are the simplest deterrent — a semi-transparent "© Your Brand" stamped across your work. But doing it manually in Photoshop for every image doesn't scale. Building a server-side Sharp pipeline from scratch is overkill for what should be a POST request.
This guide walks through the Agent Media Tools watermark API — one endpoint that takes a public image URL and your text, returns a watermarked version. No signup for basic use, no servers to deploy, no image processing libraries to install.
The watermark_image endpoint takes four inputs and returns a processed image:
| Parameter | Type | Description |
|---|---|---|
image_url | URL (required) | Public URL of the source image |
text | string (required) | Watermark text to overlay |
position | enum | One of: bottom-right, bottom-left, top-right, top-left, center. Default: bottom-right |
opacity | float 0.1–1 | How visible the watermark is. Lower = more subtle. Default: 0.5 |
The API processes the image server-side and returns the watermarked result as an image download — you save it, display it, or pass it to the next step in your pipeline.
The fastest way to try it:
curl -X POST https://agentmediatools.com/api/tools/watermark \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://example.com/photo.jpg",
"text": "© Agent Media Tools",
"position": "bottom-right",
"opacity": 0.6
}' \
--output watermarked.jpg
That's it. One command. The response is the image itself — raw binary, ready to save. No JSON wrapper, no base64 decoding, no extra steps.
If you want to see what it does with minimal effort, try it on any public image:
curl -s -X POST https://agentmediatools.com/api/tools/watermark \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://upload.wikimedia.org/wikipedia/commons/4/47/PNG_transparency_demonstration_1.png",
"text": "SAMPLE",
"position": "center",
"opacity": 0.3
}' \
--output demo-watermarked.png
For batch jobs, use Python with requests:
import requests
def watermark_image(image_url, text, position="bottom-right", opacity=0.5):
"""Add a text watermark to an image and return the bytes."""
resp = requests.post(
"https://agentmediatools.com/api/tools/watermark",
json={
"image_url": image_url,
"text": text,
"position": position,
"opacity": opacity,
},
)
resp.raise_for_status()
return resp.content
# Single image
bytes_out = watermark_image(
"https://example.com/portfolio.jpg",
"© Jane Doe Photography",
position="bottom-left",
opacity=0.4,
)
with open("portfolio-watermarked.jpg", "wb") as f:
f.write(bytes_out)
Where this really shines is processing a folder of images without looping through local PIL or Sharp calls:
import os
import requests
def watermark_batch(image_urls, text, position="bottom-right", opacity=0.5, out_dir="output"):
os.makedirs(out_dir, exist_ok=True)
for url in image_urls:
filename = os.path.basename(url.split("?")[0]) or "image.jpg"
out_path = os.path.join(out_dir, filename)
resp = requests.post(
"https://agentmediatools.com/api/tools/watermark",
json={"image_url": url, "text": text, "position": position, "opacity": opacity},
)
if resp.status_code == 200:
with open(out_path, "wb") as f:
f.write(resp.content)
print(f"✓ {filename} → {out_path}")
else:
print(f"✗ {filename} failed ({resp.status_code})")
# Example: watermark a gallery
gallery = [
"https://your-cdn.com/gallery/sunset.jpg",
"https://your-cdn.com/gallery/portrait.jpg",
"https://your-cdn.com/gallery/landscape.jpg",
]
watermark_batch(gallery, "© YourBrand 2026", opacity=0.4)
This pattern works for any number of images — just feed the URLs and let the API handle the processing on the server side. Your machine only does HTTP requests and file writes.
Watermark placement matters for both aesthetics and theft deterrence:
| Position | Best for | Notes |
|---|---|---|
bottom-right | Default. Portraits, general photography | Easy to crop out if someone's determined |
center | Preview images, social media cards | Hardest to remove, most intrusive |
top-left | Product photos on white backgrounds | Visible above fold, less common = less targeted |
bottom-left | Horizontal compositions | Balances right-side visual weight |
top-right | Images with text overlays already on bottom | Avatars and profile images |
Opacity is a trade-off. 0.3–0.5 is the sweet spot — visible enough to claim ownership, transparent enough not to ruin the image. Below 0.2 becomes invisible to most viewers. Above 0.8 dominates the photo.
If you're watermarking images for client proofs, use center at 0.5–0.7. For public social media posts, bottom-right at 0.3–0.4 works better — it's there if someone looks, but doesn't distract from the content.
If you run a Claude Desktop or any MCP-compatible agent, the watermark tool is available directly. Once connected, your agent can watermark images as part of larger workflows — for example:
"Take the latest screenshots from my website, watermark them with '© MyBrand 2026' in the bottom-right at 40% opacity, and upload them to the artifact store."
The agent chains the screenshot_page → watermark_image → create_artifact tools automatically. No code to write. If you've already set up the MCP server, this works out of the box.
For custom agent frameworks, the MCP tool definition looks like:
{
"name": "watermark_image",
"description": "Add a text watermark overlay to an image",
"parameters": {
"image_url": "https://...",
"text": "© 2026 Your Brand",
"position": "bottom-right",
"opacity": 0.5
}
}
Here are a few patterns we've seen people build with this API:
Photographers upload raw shots to a CDN. A cron job or serverless function fetches each new URL, watermarks it with "© StoreName", and saves the result back to a watermarked subdirectory. The team only manages the raw uploads — watermarks are automated.
A small Node.js middleware hooks into the image upload webhook, passes the new image URL through the watermark API, and replaces the stored version. Writers never think about watermarks — they just happen.
Photographers build a gallery page where each thumbnail is watermarked with center at 0.6 opacity. Full-resolution originals stay private. When the client pays, the watermark-free originals are delivered.
An AI agent monitors your Instagram feed, downloads each new post, watermarks it with your handle, and archives it. If a repost surfaces without credit, you have a watermarked timestamped original.
The watermark endpoint is available on the Free tier — no API key required for basic use. Free tier includes daily rate limits suitable for personal projects and light automation.
For bulk processing (thousands of images), the Pro tier at $15/month raises limits significantly, adds priority processing, and unlocks API key access with usage tracking.
No hidden overage charges. All paid tiers have hard caps — you'll never get a surprise bill.
Try it right now. Paste the curl command above with your own image URL. No signup, no credit card, no account creation.
Try the Watermark API →| Approach | Time to first watermark | Ongoing maintenance | Scales to |
|---|---|---|---|
| Manual (Photoshop/GIMP) | 30 seconds per image | Manual every time | 10s of images |
| Local Sharp/PIL script | 1–2 hours (install + debug) | Library updates, font paths | 1000s of images |
| Agent Media Tools API | 1 curl command | Zero | Unlimited (tiered) |
| Dedicated SaaS | 10 min (signup + API) | Vendor management | Depends on plan |
The API approach wins on speed and zero maintenance. The only trade-off: your source image must be publicly accessible (a URL the server can fetch). For local files, upload them to any temporary hosting (S3, Imgur, Dropbox public link) and pass the URL.
The watermark API is one of 40+ tools available through Agent Media Tools. If you're building image processing pipelines, you'll also be interested in:
All available from the same API. No additional accounts needed.
Agent Media Tools provides free and paid APIs for image processing, PDF manipulation, QR codes, web utilities, and AI agent tools. Visit the homepage to explore all available tools. Have questions? Join the Discord.