All three resize an image from a chat. Pictomancer
(ai.pictomancer/image-processing on the official registry, hosted at
api.pictomancer.ai/mcp) is a metered API: six operations, billed per call, and every input is
a URL it fetches, not a file on your disk. Image
Resize API (io.github.Br0ski777/image-resize) is narrower still: one tool, resize only,
paid per call in USDC through the x402 protocol, also from a URL. Ours reads and writes local files, does
more of the small jobs a freelancer actually needs, and charges nothing per call.
| Fact | Ours | Pictomancer | Image Resize API |
|---|---|---|---|
| Tools | 12 (10 image tools plus 2 license tools) | 6 operations: analyze, resize, compress, crop, convert, ai-generated (plus a provenance feature) | 1: media_resize_image |
| Input | A local file path on your machine | A URL Pictomancer fetches; there is no local file upload path documented | A URL the server fetches; same constraint |
| Pricing | Free tier and $19 one-time Pro, or $39 for every server | Free: analyze. Paid: from $0.001 per operation after 50 free requests, billed per call, no subscription | $0.008 per call, x402 micropayment in USDC on Base, no free tier stated beyond a 402 challenge per call |
| Account or key needed | No. No account, no key | No for the first 50 requests; an API key and card after that, per its own pricing page | No signup, but a funded x402-capable wallet is required for every call |
| Watermark, thumbnails, dominant colours | All three offered | Not offered | Not offered |
| Strip EXIF and GPS metadata | image_strip_metadata, a dedicated tool | Not documented as a distinct operation | Not offered |
| Decompression bomb guard | Declared header dimensions checked before decode, refused over 10,000 px per side | Not documented | Not documented |
| Licence | MIT | Not stated on its site or registry entry | MIT |
| Works with no network | Yes, no network call of any kind | No, every call is a network request to api.pictomancer.ai | No, every call is a network request plus an on-chain payment |
Pick it when the image already lives at a URL, such as something already uploaded to a CDN or a public
bucket, and you want resize, compress, crop, convert or an AI-generated image pipeline billed by the call
with sub-30ms processing. Its own pricing page states the first 50 requests are free with no card and no
account, and analyze, a free metadata read, lets you know the cost before committing to a
paid operation. It has no watermarking, no metadata stripping as its own operation, and no path for a file
sitting on your machine: everything is fetched from a URL first.
Pick it if you already pay per call through x402 for other tools in the same marketplace and want one more: resize from a URL, output PNG, JPEG or WebP, no signup at all beyond a funded wallet. It does one job. There is no compress, crop, watermark, batch or metadata tool, and its own README says explicitly what it is not for: OCR, QR codes or screenshots, each pointed at a different tool in the same marketplace.
Pick ours when the photo is already on your machine and should stay there: no URL to fetch, no per-call
charge, no wallet, no upload of a client's product photo to prepare it for a proposal. The measured reason
image_compress only takes a quality parameter seriously for a JPEG output is
that palette quantization was tried against a PNG and made the file 2.9 times larger (115,451 bytes
against a 39,262 byte original) rather than smaller, so the server reports the real byte count instead of
pretending a knob works where it does not. Watermarking, batch thumbnails, dominant colours and metadata
stripping are also not offered by either competitor.
The image package carries automated tests covering format detection by magic bytes, the resize and crop math, decompression-bomb refusal from declared header dimensions, and the exclusive-create path reservation that stops a batch from clobbering itself under concurrent writers. Neither competitor's site or registry entry states a test count.
Ours, Claude Code:
claude mcp add image -- npx -y @theluckystrike/mcp-imagePictomancer:
claude mcp add --transport http pictomancer https://api.pictomancer.ai/mcpImage Resize API:
claude mcp add --transport http image-resize https://image-resize.api.klymax402.com/mcpExact config file paths per client are on the setup pages, and the longer walkthrough is in resizing, compressing and watermarking images from chat.
Yes. Tool names do not collide: ours are image_resize, image_compress and so on; Pictomancer's operations are named resize, compress, crop, convert. Nothing is shared between them.
Ours. Pictomancer and Image Resize API both take a URL they fetch over the network; neither documents a local file upload path. Ours reads and writes files on disk directly and makes no network call at all.
Pictomancer is free for the first 50 requests, then from $0.001 per operation, billed per call. Image Resize API is $0.008 per call in USDC via x402, no free tier beyond its 402 payment challenge. Ours has a free tier with a 4 MP size cap and batches of 5, and a one-time $19 Pro key with no subscription and no per-call charge.
No. Neither Pictomancer nor Image Resize API documents a watermark or a dedicated metadata-stripping operation. Ours offers both: image_watermark reads the shared business profile, and image_strip_metadata re-encodes from raw pixels so EXIF, GPS and XMP are never carried across.
Pictomancer's operations and pricing come from its own site at pictomancer.ai and its official MCP registry entry for ai.pictomancer/image-processing. Image Resize API's tool, price and scope come from its README at github.com/Br0ski777/image-resize-x402 and its registry entry for io.github.Br0ski777/image-resize. All read on 2026-09-04.