原型 · 预览版:收录官方注册中心里的全部服务器,每 30 分钟同步一次。评价和账户功能将在后续阶段推出。
PopNx MCP为使用 AI 的人准备的 MCP 服务器 · 用大白话讲清楚

Crabcrop

作者:dayvson版本 0.1.0i★ –i在你的电脑上运行
注册名称io.github.crabcrop/crabcrop

安装

概览

作者提供的说明(英文原文)

Sign and explain crabcrop image URLs, fetch processed-image metadata, and purge cached results.

PopNx MCP 的三大支柱

这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。

它能帮你做什么

我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。

  • 连接方式: 在你的电脑上运行
  • 传输方式: stdio
  • 分类: 设计、图片与音频
  • 你需要填写:CRABCROP_SECRET

安装

选择你使用的 AI 应用,然后复制设置。

安装之前

  • 基础检查:1 项中通过 0 项
  • 在你的电脑上运行,拥有你的用户账户的权限(文件、网络、应用)。
  • 需要你自己填写: CRABCROP_BASE, CRABCROP_SECRET

标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。

你需要

  • npm已安装 Node.js(用于 npx 命令)

你用的是哪个应用?

在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。

打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json

{
  "mcpServers": {
    "crabcrop": {
      "command": "npx",
      "args": [
        "-y",
        "@crabcrop/mcp"
      ],
      "env": {
        "CRABCROP_BASE": "your-value",
        "CRABCROP_SECRET": "your-value"
      }
    }
  }
}

在终端中运行这条命令

claude mcp add crabcrop --env CRABCROP_BASE=your-value --env CRABCROP_SECRET=your-value -- npx -y @crabcrop/mcp

运行前请先填写空着的值。

在 Cursor 中打开

或者自己复制设置

粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json

{
  "mcpServers": {
    "crabcrop": {
      "command": "npx",
      "args": [
        "-y",
        "@crabcrop/mcp"
      ],
      "env": {
        "CRABCROP_BASE": "your-value",
        "CRABCROP_SECRET": "your-value"
      }
    }
  }
}

浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。

在 VS Code 中打开

或者自己复制设置

粘贴到项目中的 .vscode/mcp.json

{
  "servers": {
    "crabcrop": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@crabcrop/mcp"
      ],
      "env": {
        "CRABCROP_BASE": "your-value",
        "CRABCROP_SECRET": "your-value"
      }
    }
  }
}

浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。

这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。

你的选择通过 Cookie 保存在这台设备上,无需账户。

需要你填写的值

  • CRABCROP_BASE env
    Base URL of the crabcrop server the tools act on (for example https://images.example.com).
  • CRABCROP_SECRET 机密 env
    The server's signing secret, used by sign_url and purge_result. Keep it in the environment, not in tool arguments.

这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。

它能做什么

它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。

基础检查i

我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日

  • ! 注意

    作者与仓库一致

    发布者(crabcrop)不是仓库所有者(dayvson);可能是其他人代为上架的

  • … 尚未检查

    仍在维护

    我们还没有从 GitHub 读取这个仓库

  • … 尚未检查

    有许可证

    我们还没有从 GitHub 读取这个仓库

  • – 不适用

    端点有响应

    它在你的电脑上运行,没有互联网端点

  • – 不适用

    使用 OAuth 登录

    在你的电脑上运行,改用本地设置

  • … 尚未检查

    工具描述中没有隐藏指令

    它在你的电脑上运行;我们只能读取互联网端点的工具列表

与“设计、图片与音频”中的其他服务器比较i

  • 星标数多于此分类中其他 2171 个服务器的 0%
  • 发布到注册中心的时间比此分类中其他 2171 个服务器的 72% 更新
  • 基础检查结果等于或优于此分类中其他 2171 个服务器的 38%

与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。

常见问题

Crabcrop 安全吗?

它在 1 项适用的基础检查中通过了 0 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。

它需要哪些密钥或代码?

你需要填写:CRABCROP_SECRET

它在我的电脑上运行还是在云端?

它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。

它还在维护吗?

GitHub 仓库最后更新于–,最新版本(0.1.0)于12天前进入注册中心。

哪些应用可以使用它?

支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。

有什么可以替代它?

同一分类中有:Figma Context MCP, Claude Real Video, Figma MCP Server

替代选择

同一分类中其他作者的服务器,按星标数排序。

如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i

Figma Context MCP

作者:GLips

Give your coding agent access to your Figma data. Implement designs in any framework in one-shot.

在你的电脑上运行★ 16,165MIT5个月前更新
安装

Claude Real Video

作者:HUANGCHIHHUNGLeo

Let any LLM watch a video locally — and search everything it has ever watched.

在你的电脑上运行★ 2,284MIT2个月前更新
安装

Figma MCP Server

作者:figma

The Figma MCP server brings Figma design context directly into your AI workflow.

在云端运行★ 2,15213个月前更新
安装

作者的 README

打开 README

按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。

crabcrop

Website · Documentation · Playground

Blazing fast, on-demand images.

crabcrop is a Rust-built, lightning-fast on-demand image API supporting all major formats in and out, with over 50 filters and AI features like background removal. It resizes, crops, converts, and filters your images straight from a signed URL. One binary, no pre-generated thumbnails, no build step: the URL is the API.

Everything an image pipeline does, in one signed URL: resize, crop, adjust, shape, overlay, and encode, chained in the order written, plus background removal, studio mode, smart face-safe crops, redaction, and content provenance. Ask for JSON instead of an image and the meta endpoint describes the picture: dimensions, detections, OCR text, and more.

crabcrop is for sites and apps that serve the same images at many sizes and want signed, cacheable image URLs in front of an origin image store.

Install

Shell installer (macOS and Linux):

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/dayvson/crabcrop/releases/latest/download/crabcrop-installer.sh | sh

Docker: the base image, and the AI variant with background removal and face detection on board:

docker pull ghcr.io/dayvson/crabcrop:0.9.9

docker pull ghcr.io/dayvson/crabcrop-ai:0.9.9 docker run -e CRABCROP_SECRET=your-secret-change-me -p 3000:3000 ghcr.io/dayvson/crabcrop:0.9.9

A secret of at least 16 bytes is the only required configuration; the server will not start without one. By default the loader fetches public sources and refuses sources that resolve to private addresses, and an allowlist names any exceptions. See Configuration.

Release binaries: prebuilt tarballs for macOS and Linux, on x86_64 and arm64, are on the releases page.

crates.io: cargo install crabcrop builds the current release from source; the prerequisites are in Deployment.

Quickstart

1. Start the server. For local development, also allow unsigned URLs:

CRABCROP_SECRET=local-dev-secret CRABCROP_ALLOW_UNSAFE=1 cargo run --release

crabcrop listening on port 3000

2. Request an image. This resizes a source photo to 300x200:

curl "http://localhost:3000/unsafe/300x200/https://example.com/photo.jpg" --output photo.jpg

Replace example.com/photo.jpg with the URL of an image you control. You get back a 300x200 JPEG.

3. Sign your URLs in production. Drop CRABCROP_ALLOW_UNSAFE and put an HMAC signature in the first path segment:

curl "http://localhost:3000/9e24ae8bf61c0d40cb6f1ca9f9e1d3e771f6c8bc8b8ab7270e6d15458d4c14ab/300x200/https://example.com/photo.jpg" --output photo.jpg

That signature is HMAC-SHA256 of 300x200/https://example.com/photo.jpg under the secret from step 1. The SDKs and the crabcrop sign command compute it for you; the full signing rules are in URL grammar and signing.

How requests work

Every image request has three parts:

GET /<signature>/<operations>/<source-url>

The signature is an HMAC-SHA256 (hex) of everything after it, or the literal unsafe on a server that allows it. Operations are zero or more segments, applied in path order. The source is a full URL, a bucket or file locator, or a short root: reference the server expands against origins it keeps to itself.

A signed URL names a transformation of a source locator, not the source's contents: the cache key is built from the operations and the locator, so if the image at a source changes, a purged entry re-renders the new image under the same URL. Because successful responses carry a one year immutable cache lifetime, a source that must stay fixed under its URL should be versioned or immutable itself (a content hash in the source path is the usual way). A render is cached under the canonical form of its operations, so equivalent spellings share one entry. A preset: segment names an operation string held on the server, so templates stay short and the transformation menu stays operator controlled.

One URL, decoded

https://img.example.com/<signature>/crop:1200x630:smart/filters:format(webp):quality(80)/https://example.com/photo.jpg

Smart-crops the photo to the 1200x630 social card size, keeping the subject and any detected faces whole, then encodes it as WebP at quality 80. Sign it with the TypeScript or Python SDK, the crabcrop-url Rust crate, the CLI, or plain curl: same inputs, same URL.

Send Accept: application/json with the same URL and the answer is the render's meta document instead of its bytes: dimensions, format, byte size, and provenance, with richer annotations as opt-ins.

Features

  • All major formats, in and out. JPEG, PNG, WebP, GIF, AVIF, and JPEG XL, in and out. HEIC and HEIF decode only; SVG decodes, sanitized and rasterized. Name no format and the Accept header decides: AVIF first, then JPEG XL, then WebP. An explicit format(...) always wins.
  • More than photos. The source can be one frame of a video, one page of a PDF, an object in S3 or GCS, a local file, or a short root: reference. Originals can also arrive through the signed upload endpoint, off by default.
  • 51 filters, one chain. Resize, crop, fit-in, and pad are their own path segments. The 51 filters chain inside one filters: segment, everything running in the order written.
  • Smart crop, face safe. The crop window finds the subject, and keeps detected faces whole instead of clipping them at the edge.
  • Studio mode. The listing shot in one filter: cut out, backdrop, contact shadow, margin. Or studio(scene) reframes the photo on its subject and corrects the light and color.
  • Background removal. Your ONNX model, in process. The subject stays, the background becomes real alpha.
  • AI upscale. Super-resolution from a model you supply rebuilds detail instead of stretching pixels, and it never falls back to a plain resize.
  • Object removal. Name a rectangle and an inpainting model fills it in from the surrounding pixels. Everything outside the rectangle comes back bit-identical.
  • Face detection and redaction. Blur or pixelate faces, or every detected object of a class, following the object's silhouette when a segmentation model is configured. Redaction never guesses: a detection failure fails the request.
  • Metadata that describes the picture. JSON meta for every render: dimensions, format, byte size, provenance. Richer answers are opt-ins with their own model sets: CLIP tags and alt text, detections, moderation verdicts, embeddings, an aesthetic score.
  • Reads the text in the picture. Opt a meta request in and the answer reads the picture: the text lines in reading order, each with its box and a confidence score. The annotation never touches the image bytes.
  • Provenance, checked and signed. Meta reports the C2PA Content Credentials a source carries: manifest present, whether it validates, who signed it. With your own certificate, outputs are signed as served, the source recorded as an ingredient, so the chain survives the transform.
  • Sealed URLs. For links that stay public but unreadable: the whole recipe, operations and source, encrypted into one opaque token in the signature slot. Only the server opens it. The readable form stays the default.
  • Presets and purge. Server-side preset names keep templates short. Purge one result, or a whole preset group, with signatures from their own subkey.
  • URLs that explain themselves. Put explain/ in front of any signed URL and the server reports on it instead of serving the image.
  • Sources on your terms. The loader fetches only what your policy allows: host allowlists, private address blocking, the source policy re-checked on every redirect hop, and fixed fetch and decode bounds, so a small file cannot expand into gigabytes.

Documentation

This README is the front door. The full reference lives on the website and in this repo's docs folder:

  • Full documentation: every operation with examples and interactive demos, section by section.
  • Playground: build, sign, preview, and debug URLs in the browser. The secret never leaves the page.
  • The reference, in this repo:
  • URL grammar and signing: request anatomy, signing, sealed and expiring URLs, source images and source roots.
  • Operations: resize, fit-in, pad, the crop family, focal forms, debug view, presets.
  • Filters: the pixel filters, encoding directives, watermark, overlay, text, the AI filters, and studio mode.
  • Media sources: format negotiation, video output, video frames, PDF pages.
  • Meta and provenance: meta responses and C2PA verification and signing.
  • Responses and errors: response headers, the cache tiers, color management, the fallback image, and the error table.
  • Endpoints and CLI: health, metrics, purge, uploads, self-explaining URLs, cache groups, and the command line verbs.
  • Configuration: every environment variable, server-wide default operations, rotating the secret.
  • AI features: the model-backed features and declared detectors.
  • Clients and SDKs: the Rust crate, the TypeScript and Python SDKs, framework loaders, the WordPress plugin, and the MCP server.
  • Deployment: platform recipes, CDN guides, and production notes.

SDKs and integrations

Build and sign URLs from code: @crabcrop/url for TypeScript, crabcrop_url for Python, and the crabcrop-url Rust crate, all running the same spec test vectors, so client and server cannot drift. Loaders and integrations cover Next.js, Nuxt, Hydrogen, Astro, and SvelteKit, plus a WordPress plugin. For coding agents, a first-party MCP server exposes signing, explaining, meta, and purge over the same surface. Details in Clients and SDKs.

The server binary doubles as its own operator CLI: crabcrop sign, explain, meta, purge, seal, upload, and convert-thumbor run against a server (or offline, for the grammar-only verbs) with the same configuration conventions.

Migrating from another service: guides for Thumbor, imgproxy, Cloudinary, imgix, Cloudflare Images, ImageKit, and Gumlet.

Deployment

Every recipe runs the same server, configured entirely through the environment. Release images are published to GHCR as ghcr.io/dayvson/crabcrop, with the AI variant at ghcr.io/dayvson/crabcrop-ai. Recipes ship for Kubernetes (Helm and Kustomize), ECS Fargate, Cloud Run, Container Apps, Docker Compose, and a binary on a VM, plus CDN guides for Cloudflare, CloudFront, Fastly, BunnyCDN, and Akamai. See Deployment.

Development

cargo build --locked

bash scripts/test_all.sh

Security

Found a security vulnerability? Please report it privately: see SECURITY.md. Please do not open a public issue, pull request, or discussion for security problems.

License

MIT. See LICENSE.

问 PopNx