Practical overview
When it fits
Use it when the coding workflow is centered on Codex but the desired model is available through ChatGPT Web, especially when task-connected tools or MCP access are needed.
Who it is for
- Codex users who want to select ChatGPT Web models within Codex.
- Developers working on coding tasks that need task files, terminal access or tools.
- Users seeking browser-only or MCP-based full-harness workflows.
Documented uses
- Select ChatGPT Web models from Codex’s native model picker.
- Keep ChatGPT conversations attached to Codex tasks as context grows.
- Connect model tool calls to task files, terminal access and approvals through MCP.
- Use images and streaming within the Codex workflow.
Requirements & limits
The documented repository provides launcher artifacts for Windows x64, macOS Apple silicon, macOS Intel and Linux x64. Model availability depends on the ChatGPT account, and Full harness access also depends on ChatGPT workspace and administrator policy.
- Choose Browser-only mode when automatic messaging is needed without local Codex tools.
- Choose Full harness when ChatGPT tool calls must connect to the current Codex task through MCP.
- Choose Zero Risk when prompts should be pasted and sent manually through the separate MCP connector.
- Check the account’s exposed models and reasoning controls because the launcher detects available model entries.
- Check workspace policy before relying on write or modify actions.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install the launcher and add web models
Set up the launcher, sign in to ChatGPT, and make supported web models available in Codex.
Before you start
- A supported Windows, macOS or Linux release artifact.
- A ChatGPT account with models exposed to that account.
- Codex installed for the model integration.
Steps
Install the launcher using the download for the system and architecture.
Sign in to ChatGPT in the embedded browser and run the browser smoke test.
Install models and restart Codex once.
Choose a model ending in (Web) in automatic mode.
Expected outcome
Supported ChatGPT Web models appear as selectable web entries in Codex.
Guide limitations
- Available model entries depend on what the ChatGPT account exposes.
- This recipe does not configure local coding tools; Full harness setup is a separate stage.
Connect Full harness through MCP
Attach ChatGPT tool calls to the current Codex task through the documented MCP tunnel workflow.
Before you start
- The launcher and ChatGPT sign-in setup completed.
- ChatGPT Developer Mode available for the workspace.
- A ChatGPT workspace whose administrator policy permits write and modify actions.
Steps
Open MCP in the launcher, create the Tunnel and regular API key, and press Connect harness.
Create a ChatGPT Developer Mode Tunnel connector named exactly Codex Native2, with Authentication set to None and Allow all actions enabled.
Run Verify runtime to confirm that Codex Native2 is attached and available.
Expected outcome
Full harness mode connects ChatGPT to the current Codex task through MCP for files, terminal, tools and approvals.
Guide limitations
- Write and modify actions depend on workspace and administrator policy.
- Unexpected approval prompts fail closed unless the documented auto-approval option is explicitly enabled.
- The tunnel is outbound and does not expose a public IP or require router forwarding.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 1 October 2026
- Original website title
- miuuyy/codex-chatgpt-web · Official homepage title · retained in its source language