Codex Desktop
For beginners on macOS or Windows. Configure it by editing a
config.tomlfile from Settings β you can ask an AI to help draft it, no need to know TOML syntax. Codex Desktop will use GreenNode's self-hosted GLM 5.2 model.
First, complete the Prerequisites: an ACTIVE API key, the Base URL, and the GLM 5.2 model ENABLED. This page only covers install and configuration.
You'll need these 3 values (from the Prerequisites page):
Base URL (OpenAI standard)
https://maas-llm-aiplatform-hcm.api.vngcloud.vn/v1 (with /v1)
Base URL (Token Plan / package key)
https://tokenplan.api.greennode.ai/v1 β see Token Plan
API key
yourvn-... key
Model ID
z-ai/glm-5.2
GLM 5.2 is just an example model. GreenNode offers many models β swap in whichever one you want. Each model's Model ID and Base URL are on its model detail page.
Step 1 β Download and install Codex Desktop
Go to openai.com/index/introducing-the-codex-app and download the app for your machine, then install it like a normal app.
Step 2 β Open the app and log in
Open Codex and log in with your ChatGPT/OpenAI account.
Step 3 β Open the config.toml file
Click your avatar/account name in the bottom-left corner β Settings.
In the left search box, type "config.toml".
Click Open config.toml in the top-right of the Custom config.toml settings section β the file opens in your machine's default editor.
Step 4 β Add the self-hosted model configuration
Add the following block to the end of config.toml (keep the existing content above it):
Key fields explained:
model_providers.vngcloud-glm
A provider name you choose β reused inmodel_provider below
base_url
The OpenAI-standard Base URL,with /v1
experimental_bearer_token
Your API key, pasted directly into the file β Codex Desktop doesn't need an exported env var like the CLI
wire_api
Keep as"responses" β the Responses API standard Codex Desktop uses
stream_idle_timeout_ms
Max idle time (ms) before the stream is treated as timed out
request_max_retries
Number of retries on a failed request
supports_websockets
Keep as false β MaaS doesn't support websockets yet
profiles.glm
A profile name you choose β appears in the app's model picker
model
The Model ID sent to MaaS
model_provider
Points back to the provider declared above
model_context_window
Set manually since MaaS doesn't expose model metadata
model_auto_compact_token_limit
Token threshold at which Codex auto-compacts the context
model_reasoning_effort
Default reasoning effort (low / medium / high)
model_reasoning_summary
Keep as "auto" β lets Codex decide whether to summarize its reasoning
Don't copy the environment-variable style config used for Claude Code (export ANTHROPIC_BASE_URL=..., claude --model ...) β that syntax is specific to Claude Code CLI and doesn't apply to Codex. Codex Desktop reads its config from config.toml using model_providers / profiles as shown above.
Not comfortable with TOML syntax? Copy the full content of your current config.toml, paste it into an AI chat (Codex, ChatGPT, Claude...) along with the 3 values above (Base URL / API key / Model ID), and ask it to write the [model_providers.*] and [profiles.*] blocks in the correct Codex format, then paste the result back into the file.
Step 5 β Save the file and restart Codex
Save
config.toml.Fully quit and reopen Codex Desktop.
Step 6 β Verify
In the chat panel, click the model picker (bottom-right corner, e.g. "5.6 Terra Medium").
Find the profile you just added (e.g. glm) in the list β select it.
Type a test message, e.g. "Write a function to add two numbers in Python." If it responds, you're set.
Check the AI Platform Console to see the call logged.
Note on switching between the default model and your self-hosted model: If you pick a default Codex model directly from the model picker (UI), the app only overwrites the model field in config.toml β it does not reset model_provider back to the default (openai) provider. If you later want to switch back to the self-hosted model, selecting it through the UI can leave model and model_provider out of sync (pointing at mismatched providers). To be safe, whenever you switch between the default model and your self-hosted model, edit model / model_provider directly in config.toml (or ask an AI to edit it for you) instead of just switching through the model picker.
Troubleshooting
New profile doesn't show in the model picker
App wasn't restarted, or the[profiles.*] section name is wrong
Fully quit and reopen Codex; recheck the TOML syntax
401 / "Unauthorized"
Wrong or not-yet-ACTIVE API key
Recheck your key; wait forACTIVE status
404 / "Not Found"
Wrong Base URL (missing/v1)
Should be exactlyhttps://maas-llm-aiplatform-hcm.api.vngcloud.vn/v1
App errors out / can't read config
Invalid TOML syntax (missing quotes, wrong indentation)
Ask an AI to check the block you added, or compare it against the Step 4 sample
No response even with the right model selected
Out of credit, model auto-disabled
Top up credit in the AI Platform Console
Switched models via the model picker (UI) and the self-hosted model stopped working
model and model_provider in config.toml are out of sync β the picker only overwrites model, not model_provider
Reopenconfig.toml, fix model / model_provider back to a matching pair (see the Step 4 sample), save, and restart Codex
Use the CLI version
See the prerequisites
View usage & billing
Need help?
If you've followed the steps and it's still not working, feel free to contact GreenNode Customer Support:
Email: support@greennode.ai
Hotline: 19001549
Help center: helpdesk.greennode.ai
Thank you for using GreenNode's services.
Last updated