Problem connecting Azure Foundry models to VS Code Codex

Question

Problem connecting Azure Foundry models to VS Code Codex

Joel 20

We are trying to set up a foundry model to use for codex in VS Code and we're experiencing some issues, that I was hoping you could help us solve.

We then followed this guide to set it up in vscode: https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/codex?tabs=npm#use-codex-in-visual-st…

But in VS Code we still get this as a result: "unexpected status 404"

How could we resolve this please?

Additionally, we were trying to verify the cost management with our admin's account and we can see that there is an active subscription for his account but past that we cannot see whether the MfS package is active. Where exactly can we verify this?

Thank you!

Divyesh Govaerdhanan 10,890 Reputation points MVP Volunteer Moderator

2026-04-27T13:41:49.1066667+00:00
Hello Joel,

Welcome to Microsoft Q&A,

The 404 error almost always points to one of these three root causes.

Deployment name mismatch

The model field in your "~/.codex/config.toml" must match the deployment name you gave the model inside Azure AI Foundry, not the base model name.

Wrong endpoint format

The endpoint in your config.toml must use the cognitiveservices URL pattern, not the AI Services pattern.

Codex models require the Responses API, not chat/completions

The endpoint in your config.toml must use the cognitiveservices URL pattern, not the AI Services pattern.

If you're on Windows, run this inside WSL2. There are known issues running Codex CLI natively on Windows outside of WSL.

To confirm the model tier (for example, Standard vs. Global Standard) on your admin's account:

In the Azure portal, open the associated Azure OpenAI or Azure AI Services resource and go to "Model deployments." Each deployment will show its SKU tier (Standard, GlobalStandard, etc.) under the deployment details.

For billing confirmation, open Azure Cost Management + Billing, filter by the resource group containing the AI Foundry project, and look for charges under the CognitiveServices or Microsoft.MachineLearningServices resource provider. If the meter shows usage against your deployment, the subscription is active.

Please Upvote and accept the answer if it helps!!
SRILAKSHMI C 18,225 Reputation points Microsoft External Staff Moderator

2026-04-29T11:23:30.95+00:00

Hi @Joel

Did you get any chance to review the above response. Do let me know if you have any further queries.

Thank you!
SRILAKSHMI C 18,225 Reputation points Microsoft External Staff Moderator

2026-04-30T12:55:24.3433333+00:00

Hi @Joel

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!
Manas Mohanty 16,755 Reputation points Microsoft External Staff Moderator

2026-05-06T18:10:51.2566667+00:00

Hi @Joel

Yes, agree 404 relates to endpoint, model name, or api key malformed

But could you please your teams email address in private message.

Facing some issue here replicating at myside too.

Thank you.
Joel 20 Reputation points

2026-05-08T14:05:04.9133333+00:00

Hi @SRILAKSHMI C @Manas Mohanty @Divyesh Govaerdhanan

Thank you for your help!
For me the issue was that the model name was not the same as in Foundry.

Answer accepted by question author

0 additional answers

Your answer

Divyesh Govaerdhanan 10,890 Reputation points MVP Volunteer Moderator

2026-04-27T13:41:49.1066667+00:00

Hello Joel,

Welcome to Microsoft Q&A,

The 404 error almost always points to one of these three root causes.

Deployment name mismatch

The model field in your "~/.codex/config.toml" must match the deployment name you gave the model inside Azure AI Foundry, not the base model name.

Wrong endpoint format

The endpoint in your config.toml must use the cognitiveservices URL pattern, not the AI Services pattern.

Codex models require the Responses API, not chat/completions

The endpoint in your config.toml must use the cognitiveservices URL pattern, not the AI Services pattern.

If you're on Windows, run this inside WSL2. There are known issues running Codex CLI natively on Windows outside of WSL.

To confirm the model tier (for example, Standard vs. Global Standard) on your admin's account:

In the Azure portal, open the associated Azure OpenAI or Azure AI Services resource and go to "Model deployments." Each deployment will show its SKU tier (Standard, GlobalStandard, etc.) under the deployment details.

For billing confirmation, open Azure Cost Management + Billing, filter by the resource group containing the AI Foundry project, and look for charges under the CognitiveServices or Microsoft.MachineLearningServices resource provider. If the meter shows usage against your deployment, the subscription is active.

Please Upvote and accept the answer if it helps!!
SRILAKSHMI C 18,225 Reputation points Microsoft External Staff Moderator

2026-04-29T11:23:30.95+00:00

Hi @Joel

Did you get any chance to review the above response. Do let me know if you have any further queries.

Thank you!
SRILAKSHMI C 18,225 Reputation points Microsoft External Staff Moderator

2026-04-30T12:55:24.3433333+00:00

Hi @Joel

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!
Manas Mohanty 16,755 Reputation points Microsoft External Staff Moderator

2026-05-06T18:10:51.2566667+00:00

Hi @Joel

Yes, agree 404 relates to endpoint, model name, or api key malformed

But could you please your teams email address in private message.

Facing some issue here replicating at myside too.

Thank you.
Joel 20 Reputation points

2026-05-08T14:05:04.9133333+00:00

Hi @SRILAKSHMI C @Manas Mohanty @Divyesh Govaerdhanan

Thank you for your help!
For me the issue was that the model name was not the same as in Foundry.

Answer 1

Hello Joel,

Thank you for reaching out. It looks like you’re encountering two separate issues: one related to connecting Azure AI Foundry models to Codex in VS Code, and another regarding verification of your Microsoft Foundry Services (MfS) subscription.

1. Resolving the unexpected status 404 Error in VS Code Codex

A 404 Not Found error typically indicates that Codex is unable to locate your Azure AI Foundry endpoint. In most cases, this is caused by an incorrect endpoint URL or configuration.

Verify Your Codex Configuration

Please ensure your ~/.codex/config.toml file is configured as follows:

model = "YOUR_DEPLOYMENT_NAME"

Important Validation Points

The base_url must include:

Your Azure OpenAI resource name
The .openai.azure.com domain
The /openai/v1 suffix

For example:

https://myresource.openai.azure.com/openai/v1

A missing /openai/v1 path is one of the most common causes of this exact 404 error.

Additionally, please confirm that:

model exactly matches your Azure deployment name (not the underlying base model name)
model_provider = "azure" is specified
wire_api = "responses" is present
env_key contains the environment variable name, not the API key itself

2. Set the Required Environment Variables

Before launching VS Code, export the API key from the same terminal session.

macOS / Linux / WSL

export AZURE_OPENAI_API_KEY="<your-api-key>"

Setting both variables is recommended, as the Codex CLI typically uses AZURE_OPENAI_API_KEY, while the VS Code extension may reference OPENAI_API_KEY.

Windows + WSL

If you are using WSL, please also define the same environment variables on the Windows host, as the VS Code extension may read variables from the Windows environment rather than the WSL session.

3. Launch VS Code from the Same Terminal

Always launch VS Code using:

code .

from the same terminal where you exported the environment variables.

Launching VS Code from the Start menu, desktop shortcut, or dock may prevent the extension from inheriting those variables.

4. Validate Codex from the Command Line

Before testing in VS Code, verify that the Codex CLI itself is working:

codex --version

If the CLI works successfully, then your Azure configuration is correct and the remaining issue is likely isolated to the VS Code extension environment.

5. Confirm Model Deployment

Please ensure that you have deployed a Codex-compatible model in Azure AI Foundry, such as gpt-5-codex

Also verify that the value specified in model is the exact deployment name of that model.

Please refer this

Troubleshooting Codex CLI & VS Code errors: https://learn.microsoft.com/azure/foundry/openai/how-to/codex?tabs=npm#troubleshooting

Using Codex in VS Code: https://learn.microsoft.com/azure/foundry/openai/how-to/codex#use-codex-in-visual-studio-code

Azure AI Foundry model-not-found guidance: https://learn.microsoft.com/azure/ai-foundry/reference/region-support

Azure Cost Management + Billing overview: https://learn.microsoft.com/azure/cost-management-billing/costs

I Hope this helps. Do let me know if you have any further queries.

Thank you!

Share via

Problem connecting Azure Foundry models to VS Code Codex

0 additional answers

Your answer