🐛 mcp-chat: Azure OpenAI Models Not Working

[MattiaDellOca]

Workspace

mcp-chat

📜 Description

However, when making requests against an Azure OpenAI deployment, the plugin fails with:

mcp-chat error [openai] Request failed (404) after 180ms responseData="{\"error\":{\"code\":\"404\",\"message\": \"Resource not found\"}}"

Root cause:

My Azure OpenAI /chat/completions endpoint requires a mandatory api-version query parameter. Currently, the plugin always calls:

• /chat/completions

without any way to include api-version, which causes the 404 error.

Proposed solution:

Allow specifying query parameters when calling the /chat/completions endpoint. For example:

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxx.openai.azure.com/openai/deployments/yyy/'
      chatCompletionsQueryParameter: 'api-version=2024-10-21'
      token: 'zzz'
      model: gpt-4o-mini

Alternatively, if we only want to pass api-version:

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxx.openai.azure.com/openai/deployments/yyy/'
      apiVersion: '2024-10-21'
      token: 'zzz'
      model: gpt-4o-mini

This would make the plugin compatible with older Azure OpenAI deployments.

👍 Expected behavior

I would expect Azure OpenAI to work when using the plugin

👎 Actual Behavior with Screenshots

mcp-chat error [openai] Request failed (404) after 180ms responseData="{\"error\":{\"code\":\"404\",\"message\": \"Resource not found\"}}"

👟 Reproduction steps

1. Configure the plugin with:

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxx.openai.azure.com/openai/deployments/yyy/'
      token: 'zzz'
      model: gpt-4o-mini

2. Send a message in the chat

📃 Provide the context for the Bug.

No response

👀 Have you spent some time to check if this bug has been raised before?

• I checked and didn't find similar issue

🏢 Have you read the Code of Conduct?

• I have read the Code of Conduct

Are you willing to submit PR?

Yes I am willing to submit a PR!

[patrickwyler]

@MattiaDellOca with the following config I can connect to my openai model:

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxxxx.openai.azure.com/openai/v1'
      model: 'gpt-4.1-mini-model'
      token: '<token>'

[MattiaDellOca]

Hi @patrickwyler,
with that configuration I get the following error:

mcp-chat error [openai] Request failed (404) after 461ms responseData="{\n  \"error\": {\n    \"type\": \"invalid_request_error\",\n    \"code\": \"DeploymentNotFound\",\n    \"message\": \"The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.\"\n  }\n}"

[patrickwyler]

Hi @MattiaDellOca
Hmm... I used the deployment name as the model. Is that the problem?

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxxxx.openai.azure.com/openai/v1'
      model: 'deployment-xy-gpt-5.1-2025-11-13-model' <-- deployment name
      token: '<token>'

btw newer models like gpt-5.1 aren't working: (created a new issue #8402)

mcp-chat error [openai] Request failed (400) after 367ms responseData="{\n  \"error\": {\n    \"message\": \"Unsupported parameter: 'max_tokens' is not supported
 with this model. Use 'max_completion_tokens' instead.\",\n    \"type\": \"invalid_request_error\",\n    \"param\": \"max_tokens\",\n    \"code\": \"unsupported_parameter\"\n  }\n}"

I tested with gpt-4.1-mini which is working.

[MattiaDellOca]

Hi @patrickwyler,
I tried the following configuration and it doesn't work:

mcpChat:
  providers:
    - id: openai
      baseUrl: 'https://xxx.openai.azure.com/openai/v1'
      model: 'deployment-gpt-5.1'     <--- "gpt-5.1" is also the name of my deployment
      token: 'yyy'

Note that this could be an issue with my Azure OpenAI API version.
Unfortunately, I don't have direct access to it.

About newer models not being supported, this is a known limitation of the plugin. I fixed this in #7915, which is waiting for review.