Verified project record

miuuyy/codex-chatgpt-web

A desktop launcher that lets developers use ChatGPT Web models inside Codex while retaining the Codex interface, task context, streaming, and images, drawing on ChatGPT Web's separate usage limits instead of Codex quota.

Project overview

It connects ChatGPT to the current Codex task's files, terminal, tools, and approvals through MCP, so responses and tool calls stay tied to the ongoing Codex work rather than a detached chat window.

Project type: MCP · AI Agent · AI Coding
Use cases: Coding & Development · Automation
Deployment: Refer to project documentation
License: MIT

Best for

Developers and AI engineers who already use Codex, have a ChatGPT account with available Web models, and want to draw on ChatGPT Web usage limits instead of Codex quota.
Users who want a low-commitment entry point, such as the browser-only mode for sending messages without local Codex tools, before enabling the full harness.

Key capabilities

Full harness mode connects ChatGPT to the current Codex task's files, terminal, tools, and approvals through MCP.
Keeps images and streaming available in the Codex workflow when using ChatGPT Web models.
Offers a browser-only mode for sending messages without local Codex tools.
Provides a mode where users choose a connector, paste and send a prepared prompt, and confirm that it was sent.

Limitations and risks

ChatGPT UI changes can break the browser selectors the automation relies on.
The application uses unofficial browser automation rather than the OpenAI API, which carries breakage and account-policy risks compared to a supported API path.
GPU requirements, cost dependencies, and telemetry are not documented, so these factors should be verified before adoption.

Getting started

Install the launcher using the download for the system, sign in to ChatGPT in the embedded browser and run the browser smoke test, install models and restart Codex, then choose a model ending in (Web).
Setup difficulty is documented as medium: the launcher path requires installation, ChatGPT sign-in, a browser smoke test, model installation, and a Codex restart; full harness use additionally requires MCP and tunnel setup.

Alternatives and comparisons

An open source coding agent available as a VS Code extension, JetBrains plugin, CLI, desktop app, and SDK that works with any model provider; relevant for users who prefer API-based model access over browser automation.
A native, extensible open source AI agent with desktop, CLI, and API access that works with any LLM; an option for users wanting broader agent scope beyond the Codex-plus-ChatGPT Web pairing.
A plugin and CLI multi-agent orchestration harness with automated model routing and code editing tools; relevant for users who want multi-agent orchestration rather than a single ChatGPT Web connection inside Codex.

Project comparisons

Evidence and sources

GitHub project description: Use ChatGPT Web (including Pro) as a native model in Codex — with context, tools, streaming and images, without using Codex quota.
README: Full harness mode connects ChatGPT to the current task’s files, terminal, tools, and approvals through MCP.
README: <a href="https://github.com/miuuyy/codex-chatgpt-web/releases/download/v6.0.0/codex-web-gpt-6.0.0-win-x64.exe"><img src="assets/readme/download-windows.svg" width="224" height="64…
README: The app includes its browser and runtime. No separate Chrome, Node, or Bun installation is needed.
README: Full mode connects ChatGPT's tool calls back to the current Codex task through the official [OpenAI tunnel-client](https://github.com/openai/tunnel-client).

Source updated at: 2026-09-22T19:00:24Z

Open original GitHub project