QQ Chat Exporter: a local-only QQ history export with three install paths

Every documented access point is localhost:40653 with a console token

After login you work from a browser, and the address is the same on every documented path: `http://localhost:40653/qce`. The Windows one-click install ends with opening that address. Shell mode says the same thing with one extra detail: the console prints a one-click login link, and if you would rather paste the address yourself, the token is in the console output. Settings has a Startup entry that can open the browser automatically after login. Because the token is printed in the console output, treat that output as sensitive.

Three install paths, three different prices

Windows has the one-click installer the quick start marks as recommended: download it from Releases, run it, scan the QR code with QQ, open the page. Shell mode is the portable route for Windows, Linux, and the macOS Apple Silicon preview: take the platform archive, then run `launcher-user.bat` on Windows or `./launcher-user.sh` elsewhere. Docker is the only path that says it needs no QQ client installed at all, and it works on macOS (including Apple Silicon), Linux, and Windows. The trade is that Docker runs from a clone rather than a release archive:

git clone https://github.com/shuakami/qq-chat-exporter.git
cd qq-chat-exporter/docker
docker compose up -d

# 查看登录二维码与 Token
docker logs -f napcat-qce

All three end at a local browser page; shell mode prints the token in the console, and Docker shows it in the container log.

macOS gets an arm64 preview build and one hard prerequisite

The note on shell mode is narrow on purpose. macOS currently ships only an Apple Silicon (arm64) preview build, and you have to fully quit QQ on the machine before starting the exporter. An Intel Mac has no build offered in the quick start, and QQ must be fully quit before starting. The Docker note repeats the constraint in another shape: Apple Silicon (M1/M2/M3/M4) runs through Rosetta emulation, so the first start can be slower than a native run. Both caveats point at pages that live on the project docs site rather than in the repository, the macOS guide and the Docker deployment guide. On an Intel machine, or when closing the desktop client is not an option, read those two pages before you plan a backup run around them.

Local parsing moves the exposure from the network to the export file

The privacy claim is specific and easy to check: all reading and parsing happen on your own computer, the project has no server or data transfer, and chat records are never uploaded to any place. Accept that and the export itself becomes the sensitive artifact. Output formats are HTML, JSON, TXT, and Excel, and the images, videos, and sticker files sent in the chats can be pulled down beside the text. One JSON or Excel file is therefore a complete, machine-readable copy of a friend chat or a group history, with no server left in charge of it afterwards. The README does not mention encryption for those files, redaction of phone numbers, or a way to export a single conversation instead of a whole contact list, so the output directory is yours to protect.

Three viewer directories sit side by side in the root

The repository root holds qce-viewer/, qce-chunked-viewer/, and qce-v4-tool/ next to each other, plus vercel.json and a public/ directory at the top level. For a user this mostly shows up as a support detail: when a page misbehaves, say which directory it came from, because names like chunked suggest they do not all load the same way. Chunked in particular fits the data, since a multi-year group export is large and benefits from arriving in pieces. It also means the network panel is worth a look before you decide the parser is at fault.

One tag covers the Rust core, the local server, and the installer

The tree splits the pieces cleanly enough to navigate: qq-chat-export-core/ holds the Rust core, qq-chat-export-server/ holds the server, and installer/, uninstaller/, start-standalone.bat, and scripts/ cover packaging. plugins/, docker/, and docs/ sit alongside, with AGENTS.md and CLAUDE.md at the root for contributor instructions. Recent release tags: v6.3.0 on 2026-09-11, v6.3.1 on 2026-10-02, and v6.3.2 on 2026-10-03, with the last push on 2026-10-23. Two tags in two days is a fast train. Pin the version you tested, especially on a machine you cannot easily rebuild.

GPL-3.0, a NapCatQQ base, and a list of sibling projects

Two facts shape what you can do with the code and the output. The license is GPL-3.0, so a build you redistribute carries its obligations, which matters the moment you hand an installer to somebody else. And the project credits the NapCatQQ team for the framework underneath it. The related-projects block points at QCE2Chatlab, QQChatAnalyzer, QQ-Chat-AI-Analyzer, and napcat-qce-python, and the maintainers ask that community projects be listed by opening an issue. For someone choosing a tool, those links are a better comparison surface than any feature list, because they show what other people build once the export exists.

Editorial conclusion

Use QQ Chat Exporter when the job is a personal archive you can run offline: a QR login, a fixed local port, export formats you can open later without the tool. Pass if you need scheduled re-exports, per-conversation selection, or a service another machine can reach, because none of that appears in the documented paths. Before trusting an export holding years of group history, check what your platform can install, decide who can read the output directory, and read the GPL-3.0 terms if you plan to hand the build to someone else.

Frequently asked questions

Official sources

1. License: GPL-3.0
2. Project website
3. README
4. Releases
5. shuakami/qq-chat-exporter on GitHub