OneChat, as the name suggests, is a tool that can simultaneously converse with multiple online large-model websites. OneChat can send your question to multiple AI pages at once, helping you discover the best answers. Not only that, OneChat also supports one-click controls for showing or hiding all pages' sidebars, showing or hiding input boxes, starting new conversations, uploading files, quickly locating items in the conversation list, and more.
| Light Theme | Dark Theme |
|---|---|
![]() |
![]() |
v0.0.1 version only supports Windows platform:
- Download link: https://github.com/YiangQi/OneChat/releases/tag/v0.0.1
- Select onechat_0.0.1.zipand click to download;
- After extraction, double-click OneChat.exe to run.
- Send a question to multiple pages with one click
- Open all pages' login windows with one click
- Show or hide all pages' sidebars with one click
- Show or hide all pages' input boxes with one click
- Upload images or files with one click
- Jump to the previous or next question with one click
- Input content is synchronized to all pages' input boxes in real time
- Choose freely to send messages to the currently active page
- Sidebar shows the current AI conversation history
- Pages can be dragged to customize position and size
The model websites supported by OneChat can be configured dynamically via JSON. The configuration directory is the online folder in the application's root directory.
The online.json file in that folder is the model configuration file; if you want to add a custom site model, you can refer to this configuration.
Currently supported models include:
- ChatGPT
- ChatGLM
- Claude
- DeepSeek
- DouBao
- Gemini
- Kimi
- Grok
- TongYi
- YuanBao
OneChat also supports adding other large-model websites dynamically; after modifying the online.json configuration, restart the application for the changes to take effect.
-
Create a new model directory under the
onlinefolder, using the format{provider}_{model}, for exampleopenai_chatgpt. The directory name format is not strictly required, but it keeps things consistent. -
The model directory must contain two files:
inject.jsandlogo.png(the model's icon). -
inject.jsshould override the interface methods frominject_template.jsto implement the injection script for operating the model website. -
Add the model configuration option in
online.jsoninonline.js, using the following format:{ "name": "ChatGPT", "url": "https://chat.openai.com/", "icon": "openai_chatgpt/logo.png", "script": "openai_chatgpt/inject.js" }
Where:
nameis the model's display name shown in OneChat's sidebarurlis the model website's conversation addressiconis the path to the model's logoscriptis the path to the model injection script
Due to limited personal time, contributions of new model site configurations are welcome.
OneChat does not read any of your personal information. All login and personal data for model websites are stored in the tool's webcache folder in the application's root directory. If needed, you can manually delete that folder to clear personal data.
Inside the webcache directory there is a WebView folder which stores data in separate subfolders named by each model website's domain; you can delete the specific site's folder under WebView to reset that website's data.
-
The OneChat client code is built with Qt 5.15.2 and MSVC2019_64.
-
OneChat uses the third-party library QCefView as the browser engine and includes some adjustments for certain QCefView features, using CEF version 122. You can refer to the forked repository here: https://github.com/YiangQi/QCefView/tree/onechat (branch
onechat). Theqcefviewunder3rdpartyin OneChat was built from that fork. -
During runtime, you need to place the corresponding CefView module in the debug/release directory. The module folder must be named "CefView" (case-sensitive). You can directly copy it from GitHub:
- windows x64 debug cefview:https://github.com/YiangQi/OneChat/releases/tag/v0.0.1/CefView_win_x64_debug.zip
- windows x64 release cefview:https://github.com/YiangQi/OneChat/releases/tag/v0.0.1/CefView_win_x64_release.zip
-
If you want to run the release version, please first copy the build/debug/onlinedirectory to the build/releasedirectory; otherwise, the website list will be missing at runtime.

