# qaiu/netdisk-fast-download: a Java service that turns Chinese netdisk share links into direct download URLs > It aggregates share-link parsers for Lanzou, 123Pan, Quark, UC and more behind one HTTP API. Useful if you run a download site or need scriptable access to a share link, but a large share of the supported drives need cookies, and the public demo is not something to build on. **qaiu/netdisk-fast-download** — 聚合多种主流网盘的直链解析下载服务, 一键解析下载，已支持夸克网盘/uc网盘/蓝奏云/蓝奏优享/小飞机盘/123云盘等. 支持文件夹分享解析. 体验地址: https://lz.qaiu.top https://189.qaiu.top - Repository: https://github.com/qaiu/netdisk-fast-download - Website: https://nfd-parser.github.io - Stars: 3,262 · Forks: 277 - Language: Java - License: MIT - Published: 2026-09-24 · Updated: 2026-09-24 · Language: en - Canonical page: https://hysenlabs.com/projects/qaiu-netdisk-fast-download ## The problem: every netdisk speaks a different share-link dialect Chinese cloud drives do not expose a uniform download API for shared files. A Lanzou link, a 123Pan link and a mobile cloud space link each carry their own key in a different parameter, and the page in front of the file is usually JavaScript-driven. Anyone running a resource site, a mirror, an app update endpoint or a media preview page ends up writing one scraper per drive and maintaining it as the drives change. qaiu/netdisk-fast-download exists to collapse that into one HTTP call. The README describes it as a service that converts a share link into a direct link, and lists the intended consumers explicitly: download sites, resource sites, personal blogs, image hosting, app download and update flows, and video on demand. The audience is therefore developers and site operators, not end users looking for a desktop downloader. If you only ever download from one drive, a browser extension or a dedicated client is less machinery than running a Vert.x server. ## How the parser aggregation works: drive codes, share keys and one HTTP surface The architecture is a Java service built on Vert.x 4.5.27 and JDK 17 or newer, with a parser module, a web-service module and a web front end in the repository tree. Each supported drive has a short identifier used in URLs and configuration: lz for Lanzou, iz for Lanzou Youxiang, ec for mobile cloud space, fj for Xiaofeiji, ye for 123Pan, ce for self-hosted Cloudreve, and so on. The README lists roughly two dozen of these codes, including music share links from NetEase, Kugou, Kuwo and QQ Music, which are parsed by the same machinery even though they are not file drives. Two authentication paths sit on top of the parsers, and the README is explicit that they are independent. The auth query parameter carries a per-request credential: a JSON object with authType (cookie, accesstoken, authorization, password or custom) and a token field, encrypted with AES/ECB/PKCS5Padding under a 16-byte key, then Base64 encoded and URL encoded. The auths configuration is the static counterpart, set at the server level. The parser module also documents a JavaScript parser interface, with a separate guide for writing custom parsers, so a drive can be added without touching the Java core. The service listens on two ports. Port 6400 is the API and the README suggests putting Nginx in front of it. Port 6401 is the built-in web parsing tool, which the README says can be exposed directly for personal use. That split is a sensible default: the web UI has no reason to face the internet on the same port as the API. ## Installing and resolving your first share link The repository ships a Dockerfile based on eclipse-temurin:17-jre. It expects a prebuilt archive at web-service/target/netdisk-fast-download-bin.zip, unpacks it, makes run.sh executable, creates db and logs directories, and exposes port 6401. Because the build artifact is copied in rather than built inside the image, you need to produce that zip first with the Maven wrapper before running docker build. ```bash ./mvnw package docker build -t netdisk-fast-download . docker run -p 6401:6401 netdisk-fast-download ``` The README also notes that on Windows you can run the packaged build directly with run.bat, and the repository carries run.sh for Linux hosts. Once the service is up, the simplest check is the 302 redirect endpoint. Pass the share link URL-encoded, and the extraction code as pwd. The README's own example uses a mobile cloud space link with the code 1234: ```bash curl -LOJ "https://lz.qaiu.top/parser?url=https%3A%2F%2Fwww.ecpan.cn%2Fweb%2F%23%2FyunpanProxy%3Fpath%3D%252F%2523%252Fdrive%252Foutside%26data%3D70017ece572737b12b30709b7f029308eJD0%26isShare%3D1&pwd=1234" ``` Swap the host for your own and curl should follow the redirect and save the file. If you want the direct URL instead of the file, call /json/parser with the same parameters and the response is JSON rather than a redirect. There is also a short form, /d/{drive code}/{share key}@{password}, where the share key for mobile cloud space is the data parameter from the share URL. The README warns that links containing # and & must be URL-encoded before being passed as the url parameter, which is the most common way to get a confusing failure on the first attempt. ## Cookie-gated drives and the IP problem you cannot configure away The README's authentication table is the part to read before planning anything. Quark and UC are marked as requiring a cookie: without one, parsing does not work at all. Xiaofeiji and Lanzou Youxiang need authentication only for large files. 123Pan needs it for large files or for shares that require login, and accepts either an account and password or a token. The auth parameter is the mechanism, and the README points to an online tool for building the encrypted value, but the underlying constraint is that you must supply credentials you own. There is a second constraint that no configuration fixes. The README states plainly that Xiaofeiji parsing has an IP restriction and that most mainland IPs from cloud providers are blocked, and that this is unrelated to the program. A proxy is the suggested workaround. The same section warns that Xiaofeiji has been suspending accounts frequently and suggests choosing another drive for sharing. For 123Pan the README goes further and says it is only recommended for local deployment, because public and cloud servers face strict risk control, while noting that Windows users can start it with run.bat. A service that works from a laptop and fails from a VPS is a deployment property, not a bug you can patch. Support also decays. The README records that 123Pan parsing for files over 100MB broke on 2023-08-24 and now needs a login, and that UC parsing broke on 2023-07-22 and needs a login. 115 and Weiyun are struck through as no longer working. Treat the drive list as a snapshot with an expiry date attached to each entry. ## The free and professional split, and what it means for Baidu or Quark users Several drives are listed under a professional edition heading rather than in the main support list: Baidu netdisk open platform, Xunlei, Guangyapan, Quark, UC, mobile cloud disk, China Unicom cloud disk and Tianyi cloud disk. That is a meaningful boundary. If your requirement is a Baidu link, this repository's MIT-licensed code is not the thing you install; the README points to the professional edition at 189.qaiu.top and to an API documentation site. Quark and UC appear in the authentication table with a mandatory cookie requirement, which is consistent with them being professional-edition targets. The README does not describe the professional edition's licence, pricing or deployment model, so anyone whose use case depends on those drives has to resolve that question outside this repository. For the free set, the licence is MIT, which is permissive and carries no copyleft obligation. Note that the repository also carries a core-database module and a db directory in the Docker image, so a self-hosted instance keeps local state; the README does not document a migration or backup procedure for it. ## Alternatives: purpose-built tools versus a self-hosted resolver The closest alternatives are per-drive downloaders and browser userscripts. Tools in the QuarkPanTool and 123Pan downloader family target one drive each, run on your desktop, and reuse your browser session, which sidesteps the cookie encryption dance entirely. The difference in approach is deployment shape: those are clients, this is a server. A client cannot answer an HTTP request from your blog, cannot resolve a folder share into a file list for a web page, and cannot be called from a build script on another machine. Baiduwp occupies a similar niche for Baidu links specifically, and it is worth noting that Baidu is not in this project's free support list at all, so for that drive the two are not really comparable. A generic scraping framework is the other option: you write the parsers yourself and keep full control, at the cost of owning the maintenance for every drive you support. This project's value proposition is that someone else already wrote those parsers and keeps the drive codes, share-key extraction and folder listing in one place. If you only need one drive and you are willing to maintain one scraper, that trade is worse for you. ## Maintenance cadence and upgrade cost The last push to the repository was on 2026-09-19, the same day as the v0.4.6 release, and v0.4.5 and v0.4.4 landed a month earlier. That is a steady release rhythm, and the drive list in the README carries dated entries showing when individual drives broke, which suggests breakage is tracked rather than ignored. The flip side is that this cadence is driven by external change: when a drive alters its share page, a release is needed. Upgrading is therefore not optional maintenance you can defer for a year unless you freeze your drive list too. The upgrade path itself is a Maven build plus a redeploy of the unpacked archive or the Docker image. The README does not document rollback, database schema migrations or backward compatibility guarantees between minor versions, so pin the version you deploy and keep the previous archive. On the licence side, MIT applies to this repository; the professional edition is a separate offering whose terms the README does not state. Nothing here is legal advice, and if you plan to resell access, read the actual licence file and the professional edition terms yourself. ## Conclusion Adopt it if you control the deployment: a Java 17 host or the Docker image, your own cookies for Quark and UC, and a download site or script that needs a stable /parser endpoint instead of hand-written per-drive scraping. Do not adopt it if you need Baidu, Xunlei, Quark or UC without paying for the professional edition, or if you intend to point production traffic at lz.qaiu.top, which the README itself tells you not to over-rely on. Before committing, verify that the drives you actually care about still resolve with your own credentials in v0.4.6, and check whether the professional edition is required for the ones you need. The MIT licence covers the code in this repository, not the professional edition modules, which the README lists separately. ## FAQ ### What is qaiu/netdisk-fast-download? It is a Java service that converts cloud drive share links into direct download links, aggregating parsers for Lanzou, Lanzou Youxiang, mobile cloud space, Xiaofeiji, 123Pan, Cloudreve and others behind a single HTTP API. The README lists download sites, resource sites, blogs, image hosting and app update flows as intended uses. ### How do I install and run qaiu/netdisk-fast-download? Build the package with the Maven wrapper, then build the Docker image from the provided Dockerfile, which is based on eclipse-temurin:17-jre and exposes port 6401. The README also notes that Windows users can start the packaged build with run.bat. ### Does qaiu/netdisk-fast-download support Quark and UC drives? Quark and UC are listed under the professional edition rather than the main support list, and the authentication table marks a cookie as mandatory for both. The README does not state the professional edition's licence or pricing. ### Why does parsing fail on a cloud server? The README states that Xiaofeiji parsing has an IP restriction and that most mainland IPs from cloud providers are blocked, suggesting a proxy as the workaround. For 123Pan it recommends local deployment only, because public and cloud servers face strict risk control. ## Sources - [License: MIT](https://github.com/qaiu/netdisk-fast-download/blob/main/LICENSE) - [Project website](https://nfd-parser.github.io) - [qaiu/netdisk-fast-download on GitHub](https://github.com/qaiu/netdisk-fast-download) - [README](https://github.com/qaiu/netdisk-fast-download/blob/main/README.md) - [Releases](https://github.com/qaiu/netdisk-fast-download/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/qaiu-netdisk-fast-download