Codex
Codex can be used with the GWDG commercial model offering via Azure OpenAI.
What You Need
- a commercial endpoint and API key from support@gwdg.de
- the deployment name that should be used with your resource
- Visual Studio Code with the OpenAI extension
Codex For Visual Studio Code
1. Install the VS Code extension
- In the VS Code Marketplace, install
openai.chatgpt. - Documentation: Codex IDE extension
2. Azure OpenAI configuration
During onboarding, you may receive ready-to-use config.toml and .env files from GWDG.
If not, create the files manually with the same structure shown below.
Save both files in the Codex config directory:
- Windows:
%USERPROFILE%\.codex\ - macOS:
~/.codex/ - Linux:
~/.codex/
Target paths:
- Windows:
%USERPROFILE%\.codex\config.tomland%USERPROFILE%\.codex\.env - macOS:
~/.codex/config.tomland~/.codex/.env - Linux:
~/.codex/config.tomland~/.codex/.env
Use the exact base URL, API key, and deployment name that you received from GWDG. The default configuration should point to your provisioned Azure deployment.
- Config basics: Codex config basics
- Full reference: Codex config reference
After saving the files:
- Restart VS Code.
- Open the Codex sidebar.
3. Example config.toml
model = "<your-gwdg-deployment-name>"
model_provider = "azure"
approval_policy = "on-request"
sandbox_mode = "workspace-write"
[model_providers.azure]
name = "Azure"
base_url = "https://<your-resource>.openai.azure.com/openai"
wire_api = "responses"
query_params = { api-version = "2025-04-01-preview" }
env_key = "AZURE_OPENAI_API_KEY"
env_key_instructions = "Set AZURE_OPENAI_API_KEY in ~/.codex/.env"4. Example .env
AZURE_OPENAI_API_KEY=<your-api-key>Replace the placeholders with the values from your GWDG onboarding.
5. Sandboxing
For local work, the recommended setup is workspace-write with approvals enabled.
Codex uses local sandbox controls to restrict filesystem and network access.
- Details: Sandboxing
- Security and approvals: Agent approvals and security
Note
AI models can hallucinate and generate incorrect information. Always verify generated results critically before using or processing them.
Documentation
What You Need
- a commercial endpoint and API key from support@gwdg.de
- the deployment name that should be used with your resource
- a local terminal
npmfor Codex CLI installation
Codex CLI
1. Install the CLI
Install the Codex CLI with npm:
npm i -g @openai/codexOfficial documentation:
2. Azure OpenAI configuration
Codex CLI uses the .codex configuration directory in your home folder.
Create config.toml and .env in the following location:
- Windows:
%USERPROFILE%\.codex\ - macOS:
~/.codex/ - Linux:
~/.codex/
Target paths:
- Windows:
%USERPROFILE%\.codex\config.tomland%USERPROFILE%\.codex\.env - macOS:
~/.codex/config.tomland~/.codex/.env - Linux:
~/.codex/config.tomland~/.codex/.env
Use the exact base URL, API key, and deployment name that you received from GWDG.
3. Example config.toml
model = "<your-gwdg-deployment-name>"
model_provider = "azure"
approval_policy = "on-request"
sandbox_mode = "workspace-write"
[model_providers.azure]
name = "Azure"
base_url = "https://<your-resource>.openai.azure.com/openai"
wire_api = "responses"
query_params = { api-version = "2025-04-01-preview" }
env_key = "AZURE_OPENAI_API_KEY"
env_key_instructions = "Set AZURE_OPENAI_API_KEY in ~/.codex/.env"4. Example .env
AZURE_OPENAI_API_KEY=<your-api-key>Replace the placeholders with the values from your GWDG onboarding.
5. Start the CLI
Open a terminal in your project directory and run:
codexCodex can inspect your repository, edit files, and run commands in the current directory. If no provider is configured yet, Codex prompts for sign-in on first launch. For the GWDG commercial setup, configure the Azure provider first and then start the CLI.
6. Update the CLI
To update Codex CLI:
npm i -g @openai/codex@latest7. Sandboxing
Use workspace-write with approvals enabled unless you have a specific reason to allow more.
- Details: Sandboxing
- Security and approvals: Agent approvals and security
Note
AI models can hallucinate and generate incorrect information. Always verify generated results critically before using or processing them.
Documentation
Notes
- Codex CLI is available on macOS, Windows, and Linux.
- On Windows, OpenAI documents both native PowerShell usage and WSL2.
More AI Services
In addition to Azure OpenAI, we offer other AI services. An overview of all available AI services can be found here: