Analysis updated 2026-05-18
Connect Codex to DeepSeek's web chat instead of paying for the official DeepSeek API.
Experiment with tool-calling behavior against a browser-based DeepSeek session.
Run a local OpenAI-compatible endpoint backed by a real DeepSeek web login.
| db9cd2fgbj-hash/codex-deepseek-bridge | deanwhitex/aiw2.0stack | xmanner/miaoshou-auto-edit | |
|---|---|---|---|
| Stars | 28 | 28 | 28 |
| Language | JavaScript | JavaScript | JavaScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 3/5 |
| Audience | developer | vibe coder | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Requires Node.js 18+, a Chrome or Edge browser, and a DeepSeek web account to log in through.
Codex DeepSeek Bridge is a local server, written in JavaScript, that connects the DeepSeek web chat interface to Codex (OpenAI's coding tool). It captures a DeepSeek web login session using your browser and re-exposes it as a local API endpoint that follows the OpenAI format, so Codex can communicate with DeepSeek as if it were calling an official paid API. The project is not affiliated with DeepSeek and does not use the official DeepSeek API. The bridge uses a library called Playwright to open Chrome or Microsoft Edge, walk through the login flow at chat.deepseek.com, and save those credentials locally. Once running, it starts a server at http://127.0.0.1:8787/v1 and translates incoming OpenAI-style requests into web chat interactions, then returns the responses in OpenAI-compatible format. It bundles a Proof of Work module that DeepSeek's web interface requires internally, so no extra configuration is needed for that component. The project supports basic tool-calling features, including function calls, custom tool calls, and tool search calls, by parsing text patterns in DeepSeek's chat output. The README notes this approach is less reliable than native tool-calling APIs because it depends on matching output text rather than a formal API contract. To use it you need Node.js 18 or newer, npm, Chrome or Edge, a DeepSeek web account, and a version of Codex Desktop or Codex CLI that supports a custom model provider. Running npm run login opens a browser for you to log in, then npm start launches the local gateway. You point Codex at the local URL and set the model name to deepseek-web/deepseek-chat or deepseek-web/deepseek-reasoner. The project is not licensed for commercial use. It is available for personal learning, research, and non-commercial purposes only. Commercial use requires written permission from the author, and the author disclaims responsibility for account issues, service restrictions, or data loss.
A local tool that lets Codex talk to DeepSeek's free web chat by impersonating an OpenAI-style API, so you can use DeepSeek without the paid official API.
Mainly JavaScript. The stack also includes JavaScript, Node.js, Playwright.
Free for personal learning, research, and non-commercial use only, commercial use needs the author's written permission, and there is no warranty.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.