byondr Image Tools

npx add-mcp 'https://byondr.co/image-tools/mcp'

Resize

Change width and height of JPG, PNG, WebP and HEIC

Compress

Make photos smaller with little visible change

Convert

Switch between JPG, PNG, WebP and AVIF

Remove EXIF

Strip location and camera details before sharing

Crop

Cut images to a square, 16:9 or any size

Rotate & flip

Turn images 90°, 180° or mirror them

Watermark

Put your name, website or logo on your photos

process_images({
  file_id: "img_7Qk",
  width: 1600,
  format: "webp"
})

More image tools soon, built the same way.

Add image tools to your agent in two steps

The same seven tools, over MCP, A2A or REST. Agents upload a file or pass a URL and get back a download link and a list of the metadata removed.

1. Once per app — paste into your agent:

Read https://byondr.co/image-tools/start.md and follow it to add byondr Image Tools to this app. Ask me before changing any settings.

2. Then ask:

Use byondr Image Tools to resize, compress, convert, crop, rotate, watermark or strip metadata from my images.

Claude Code

claude mcp add --scope user --transport http byondr-image-tools https://byondr.co/image-tools/mcp

MCP config

{
  "mcpServers": {
    "byondr-image-tools": {
      "url": "https://byondr.co/image-tools/mcp"
    }
  }
}

A2A

Agent card: https://byondr.co/.well-known/agent-card.json
A2A endpoint: https://byondr.co/a2a

REST

POST https://byondr.co/image-tools/api/process_images
Send the process_images arguments as a JSON body.

How it works

Pick a task above, drop in an image or a zip of images, choose your settings and download the result. The pages run in your browser: your images are processed on your own device and are not uploaded. Only files your browser cannot handle on its own, such as iPhone HEIC photos in some browsers, AVIF output where your browser cannot create it, and very large images, are sent to our server instead. The server applies every change in one pass, in a fixed order: crop, rotate and flip, resize, watermark, then save in the format you chose. One image in gives one image out. A zip in gives a zip out. Names that would clash get a -2 suffix and unsafe names are cleaned up. Files in a zip that are not supported images, such as documents, videos and system files, are skipped and listed.

Files sent to the server are stored only while they are processed and so you can download the result. Download links expire after 24 hours, and files are deleted automatically within about 2 days.

What each option does

Resize

Set a width, a height, or both. Scale-down, the default, shrinks large images to fit and never enlarges small ones, so a batch of mixed sizes comes out no bigger than you asked. Contain fits the image inside the box and may enlarge it. Cover fills the whole box and trims what does not fit, which is the right choice for square thumbnails. Pad fits the image inside the box and fills the rest with a plain border.

For a batch of different shapes, give the longest side instead (`max_side`): every image gets that size on its long edge, portrait or landscape, and is never enlarged unless you pick contain. Or give the width and height as percentages of each image (`resize_unit: "percent"`), so 50 halves every image whatever its size.

Crop

Crop cuts out a rectangle before anything else happens. Give the left and top edge and the width and height in pixels, measured on the image as you see it: phone photos are turned upright first using their orientation tag, so the numbers match what your photo viewer shows. Agents can give the four values as percentages of each image instead (`crop_unit: "percent"`), or ask for the largest centered crop of a ratio such as 1:1 or 16:9 (`crop_aspect`); both are worked out for every image in a zip on its own, so a batch of different sizes never fails.

Compress and convert

Quality runs from 1 to 100. Around 75 is a good default for photos: files shrink a lot and the difference is hard to see. WebP is usually the best choice for websites; it is smaller than JPEG at the same quality and every current browser shows it. AVIF is smaller still but slower to create, and AVIF images larger than 1,200 pixels on the longest side are scaled down to fit here. PNG keeps every pixel exactly and suits screenshots, logos and images with transparency, but photos saved as PNG are large. Keep leaves the format alone, except that iPhone HEIC photos become JPEG and GIFs become PNG, so the result opens everywhere.

Rotate and flip

Rotate turns the image clockwise by 90, 180 or 270 degrees. Flip mirrors it horizontally, vertically or both. Photos that look sideways in some apps are already fixed automatically, because their orientation tag is applied before any other change.

Strip metadata

Photos from phones and cameras carry hidden EXIF data: the GPS position where the photo was taken, the camera or phone model, the date and time, and sometimes the owner's name. Anyone who downloads the original file can read it. With Strip metadata on, which is the default, all of it is removed from JPEG output, and WebP and PNG outputs carry none. On Remove EXIF you can keep the EXIF date taken or the EXIF copyright, and nothing else. On the Remove EXIF page, JPEG photos are cleaned without being saved again, so the picture and its colour profile stay exactly as they were (extra images, such as a depth or HDR map, are removed with the metadata). Any other change saves the image again, and its colours may be converted to sRGB, which looks the same on most screens; wide-gamut photos can look slightly less vivid.

Watermark

Add a line of text such as your name or website. Choose a corner, the centre, or tile it across the whole image, or drag it anywhere on the preview, and set its colour, size and opacity. A white watermark at about 60 percent opacity reads well on most photos; use dark text on bright images. The default size is 4 percent of the image width, so it scales with each image in a batch.

Limits

• In your browser: up to 200 images per batch.
• On our server: up to 50 images and 500 files in total per zip, 100 MB per zip and 20 MB per image.
• Up to 25 MB per file when a file is sent to our server from this page; agents can send up to 3 MB inline and upload up to 100 MB.
• Up to 12,000 pixels on a side and 100 megapixels per image.
• AVIF output is scaled down to fit 1,200 pixels on the longest side; WebP output over 36 megapixels is scaled down to fit, not rejected; PNG output up to 40 megapixels.
• Inputs: JPEG, PNG, GIF, WebP and HEIC. Animated GIF and WebP files are processed as still images.
• Download links expire after 24 hours; files are deleted automatically within about 2 days.
• If metadata cannot be removed from a JPEG, processing stops instead of returning the file.

For AI agents and developers

Image Tools is also an MCP server at https://byondr.co/image-tools/mcp with two tools. process_images takes exactly one source, a file_id, a public https url or base64 data, plus the same options as this page, and returns a download link with the size, format and byte count of every output, and the kinds of metadata each image carried and still carries (location, camera, serial number and so on; never the values). create_upload returns a one-hour, single-use upload link for files over 3 MB (up to 100 MB). The inline limit is lower than the upload limit because the MCP transport accepts request bodies of 4 MB at most and base64 adds a third. Agents call it for free within a daily quota, counted in output megapixels; a call that fails on our side does not count.

Frequently asked questions

Is Image Tools free?

Yes, on this page, with no limit for files your browser can handle. AI agents can call the MCP tool for free within a daily quota.

Do you keep my images?

Not if your browser can do the work itself: the pages process images on your device first and nothing is uploaded. Only files your browser cannot handle, such as HEIC photos in some browsers or very large images, and calls from AI agents reach our server. The server deletes them automatically within about 2 days, download links expire after 24 hours, and they are not used for anything else.

Does it remove GPS location from my photos?

Yes. With Strip metadata on (the default), EXIF data including GPS coordinates, camera model and capture time is removed from every output. Agents can keep only the date taken or the copyright with keep_date and keep_copyright.

How many images can I process at once?

In your browser, up to 200 images per batch. Files that go to our server have tighter limits: up to 50 images and 500 files in total in one zip, 100 MB per zip and 20 MB per image. Files sent from this page to the server are limited to 25 MB; agents can send up to 3 MB inline and upload up to 100 MB with create_upload.

Which formats are supported?

Input: JPEG, PNG, GIF, WebP and HEIC (iPhone photos). Output: JPEG, PNG, WebP and AVIF. AVIF images larger than 1,200 pixels on the longest side and WebP images larger than 36 megapixels are scaled down to fit.

How do AI agents use it?

Connect an MCP client to https://byondr.co/image-tools/mcp and call process_images with a file_id, a public https url or base64 data. The result is a 24-hour download link with the size and format of every image and the kinds of metadata found and removed (never the values).