codeweb.chat

What it is

Code Web Chat (CWC) is an open-source AI coding assistant for VS Code–style editors, including VS Code, Cursor, VSCodium, Antigravity, and similar environments. It does not provide its own model. Instead, it organizes your project context into coding prompts and sends them to chatbots such as ChatGPT, Gemini, Claude, Grok, DeepSeek, Kimi, and Qwen, or uses models through APIs such as the Gemini API, OpenRouter, and local Ollama. It can then apply the model’s responses back into your code workspace.

Core capabilities

CWC focuses on “context and editor workflow.” It supports selecting files, using the contents of open editors, recognizing paths, and copying context in XML format. It can also use a browser Connector extension to parse web pages into Markdown and add them to the context. Its API tools cover code completion, natural-language file editing, file creation/deletion, intelligent handling of truncated edit formats and problematic diffs, and generating commit messages that match the project style. It also provides checkpoints, allowing you to restore the workspace state after applying a response.

Pricing and privacy

The main text clearly states that CWC is free and open source under the GPL-3.0 license; no commercial edition or subscription pricing is disclosed. If you use external APIs or chatbots, costs and quotas depend on the corresponding service, but the page does not provide details. On privacy, CWC says it runs locally, with the editor and browser communicating via local WebSocket, while API tools call model providers directly. The browser extension is open source and requests relatively few permissions. However, once code is sent to a third-party model, it is still subject to that platform’s policies.

Pros, cons, and limitations

Its strengths are broad model compatibility, the ability to reuse existing chatbot accounts, and a workflow that is more suitable for real project context than simple copy-and-paste. Features valuable to developers include diff repair, code completion, commit message generation, and state restoration. The limitations are also clear: AI quality depends on external models; “Apply response” is not an automatically reliable output-extraction mechanism and still requires manual review; remote SSH scenarios require forwarding port 55155. For Chinese-language use, it supports several Chinese model entry points, but there is no visible commitment to a Chinese interface or Chinese-language support.

Who it’s for and access from China

CWC is suitable for individual developers and small teams who are comfortable with VS Code, want freedom to choose models, and value open-source controllability. The main text does not describe access from China, and some of the supported overseas chatbots/APIs may face network or payment restrictions. Possible options to consider include DeepSeek, Tongyi/Qwen, Kimi, Doubao, local Ollama models, or alternatives such as Cursor, Continue, Cline, and Aider.