# AI use cases for Codex ChatGPT Web (miuuyy/codex-chatgpt-web) Repo: [miuuyy/codex-chatgpt-web](https://github.com/miuuyy/codex-chatgpt-web) | Use cases: 4 | Page: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/ A launcher that lets you use ChatGPT Web models, including Pro, as a native model inside Codex, with optional MCP tools for local files and terminal access. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Use ChatGPT Web models inside Codex without Codex quota: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/use-chatgpt-web-models-in-codex/ - Let ChatGPT use files and terminal in a Codex task: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/let-chatgpt-use-codex-task-tools/ - Send prepared prompts to ChatGPT with Zero Risk mode: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/paste-prompts-with-zero-risk-mode/ - Set cross-backend subagent delegation in Codex: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/switch-subagent-delegation-protocol/ ## How to use ChatGPT Web models inside Codex without Codex quota with Codex ChatGPT Web Repo: [miuuyy/codex-chatgpt-web](https://github.com/miuuyy/codex-chatgpt-web) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Add ChatGPT Web models to Codex's native picker and keep Codex quota unused. **Who it is for:** A developer who already uses Codex and wants access to ChatGPT Web models, including Pro, from the same interface and task flow. **The problem:** Codex work spends Work or Codex quota, and the ChatGPT Web models you pay for are not available inside Codex's model picker. **The outcome:** Codex shows ChatGPT Web models ending in (Web) and sends messages to them with images and streaming, while using ChatGPT Web's separate usage limits. ### Steps 1. On macOS or Linux, run `curl -fsSL https://github.com/miuuyy/codex-chatgpt-web/releases/latest/download/install-launcher.sh | sh`. On Windows PowerShell, run `irm https://github.com/miuuyy/codex-chatgpt-web/releases/latest/download/install-launcher.ps1 | iex`. 2. Open the launcher, sign in to ChatGPT in the embedded browser, and run the browser smoke test. 3. Install the models from the launcher and restart Codex once. 4. In Codex, choose a model ending in (Web). For Sol models, pick the reasoning level through Effort. ### Prompt to paste into your agent ```text Using the codex-chatgpt-web repo, check that the launcher is installed and that a model ending in (Web) appears in Codex's model picker for . Report the model names you see and whether the browser smoke test passed. Verify that no Codex quota is used by running one short prompt on a (Web) model and confirming the reply arrives. ``` ### Caveat The project is unofficial browser automation, is not affiliated with OpenAI, and its release builds are not yet platform-signed, so Gatekeeper or SmartScreen may warn. Risk, credentials: The launcher keeps a signed-in ChatGPT browser profile, which is a sensitive login artifact, so use a trusted workstation and never share the launcher profile. Risk, remote script: The one-line installer downloads and runs a script from GitHub, so review it first if you do not trust the source. Risk, reverse engineering: This is unofficial browser automation and not an OpenAI API, so ChatGPT UI changes can break it. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/use-chatgpt-web-models-in-codex/ ## How to let ChatGPT use files and terminal in a Codex task with Codex ChatGPT Web Repo: [miuuyy/codex-chatgpt-web](https://github.com/miuuyy/codex-chatgpt-web) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Connect ChatGPT to a Codex task's files, terminal, and tools through MCP. **Who it is for:** A developer who wants ChatGPT models to read files, run terminal commands, and use tools inside a live Codex coding task. **The problem:** Browser-only ChatGPT cannot see your local project, so you copy code and output back and forth by hand. **The outcome:** ChatGPT tool calls reach the current Codex task through an outbound tunnel, and the Codex Native2 connector reports as available in Verify runtime. ### Steps 1. Finish the required setup in the launcher, then open MCP. 2. Create the Tunnel and a regular API key, then press Connect harness. 3. In ChatGPT, enable Developer Mode and create a new Tunnel connector named exactly Codex Native2 with Authentication: None and Allow all actions. 4. In the launcher, run Verify runtime to confirm that Codex Native2 is attached and available. ### Prompt to paste into your agent ```text Using the codex-chatgpt-web repo, guide the setup of Full harness mode for the Codex task in . Confirm the connector name is exactly Codex Native2 and that Verify runtime reports it as attached and available. Do not enable auto-approve; report any approval prompts you see. ``` ### Caveat Unexpected approval prompts fail closed unless --auto-approve-tool-calls is explicitly enabled, and that option only clicks Allow once, never a permanent grant. Risk, credentials: The setup creates a regular API key and a Tunnel, so store the key securely and revoke it if the machine is shared. Risk, reverse engineering: Write and modify actions also depend on your ChatGPT workspace and its administrator policy, and the browser integration is unofficial. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/let-chatgpt-use-codex-task-tools/ ## How to send prepared prompts to ChatGPT with Zero Risk mode with Codex ChatGPT Web Repo: [miuuyy/codex-chatgpt-web](https://github.com/miuuyy/codex-chatgpt-web) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Use a manual paste-and-send flow when you do not want the launcher to operate the ChatGPT page. **Who it is for:** A developer who wants ChatGPT to use Codex tools but does not want the launcher to read or drive the ChatGPT page. **The problem:** Browser-only and full harness modes send messages automatically, which may be more automation than you want for sensitive work. **The outcome:** You send each prepared prompt yourself in ChatGPT, and the launcher shows Sent once you confirm, while ChatGPT still reaches local Codex tools through a separate MCP connector. ### Steps 1. Install the launcher and sign in to ChatGPT in the embedded browser. 2. In Codex, choose the Zero Risk model entry. 3. Set up the Codex Zero Risk connector in ChatGPT, then choose the model and connector yourself in ChatGPT. 4. Paste the prepared prompt into ChatGPT, send it, and confirm Sent in the launcher. ### Prompt to paste into your agent ```text Using the codex-chatgpt-web repo, prepare a Zero Risk prompt for and list the exact model and Codex Zero Risk connector the user should select. Expected output is the prompt text and a checklist of manual steps. Verify that you did not send the prompt yourself and that the user confirms Sent in the launcher afterward. ``` ### Caveat Zero Risk does not read or operate the ChatGPT page, so every message is sent by you, which slows routine work. Risk, credentials: The Zero Risk connector still gives ChatGPT access to local Codex tools, so review each action before you approve it. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/paste-prompts-with-zero-risk-mode/ ## How to set cross-backend subagent delegation in Codex with Codex ChatGPT Web Repo: [miuuyy/codex-chatgpt-web](https://github.com/miuuyy/codex-chatgpt-web) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Choose between Compatibility V1 and Native subagent delegation for Codex tasks. **Who it is for:** A developer who runs Codex subagents across different model backends and needs the delegation mode to match their setup. **The problem:** Subagents across backends may fail or behave differently depending on the protocol, and it is unclear which mode a Codex task uses. **The outcome:** You know the current subagent protocol, and the task runs on the mode you chose after a Codex restart and a new task. ### Steps 1. Run `codex-chatgpt-web subagents status` to see the current protocol. 2. Run `codex-chatgpt-web subagents compatibility-v1` to use Compatibility V1, which new installs use by default. 3. Or run `codex-chatgpt-web subagents native` to keep Codex's own feature settings and enable plaintext Web-to-Web V2 delegation. 4. Restart Codex and start a new task after changing the protocol. ### Prompt to paste into your agent ```text Using the codex-chatgpt-web repo, run the subagents status command and report the current protocol. Then set the protocol to and confirm the new value by running status again. Expected output is the two status results and a note to restart Codex. ``` ### Caveat Native mode enables plaintext Web-to-Web V2 delegation, so confirm your data handling is acceptable before switching. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/miuuyy-codex-chatgpt-web/switch-subagent-delegation-protocol/