Azure OpenAI node fails with custom gateway endpoint while HTTP Request works (AI Agent incompatible)

Questions
1

Description

I am experiencing an issue when using Azure OpenAI nodes with AI Agent in n8n when a custom gateway / proxy endpoint is involved.

Environment

  • n8n: Docker deployment

  • Using AI Agent

  • Backend: Azure OpenAI

  • Access pattern: Azure OpenAI is accessed via a custom gateway domain, not directly via *.openai.azure.com

What works

  • Using HTTP Request node, the following Azure-style endpoint works correctly:

    POST /openai/deployments/{deployment}/chat/completions

  • The request succeeds with the same headers, model, and payload.

What fails

  • Using Azure OpenAI Chat Model / Azure OpenAI node (required by AI Agent) fails with:

    The resource you are requesting could not be found

  • This happens even though:

    • DNS resolution works

    • The same deployment and API key work via HTTP Request

    • The gateway correctly forwards traffic to Azure OpenAI

Observations

  • The Azure OpenAI node requires an “Instance / Resource Name” field.

  • This field appears to be treated as an Azure resource name and internally used to construct:

    https://{instance}.openai.azure.com

  • When a custom gateway domain is provided instead, the node appears to:

    • Construct an invalid endpoint, or

    • Combine instance name and endpoint incorrectly

  • As a result, the Azure OpenAI node does not support custom gateway domains, even though HTTP Request does.

Limitation Impact

  • AI Agent does not support HTTP Request as a model provider.

  • Therefore, users behind gateways/proxies cannot use Azure OpenAI with AI Agent unless:

    • The gateway fully mimics the official Azure OpenAI endpoint, or

    • They switch to OpenAI-compatible endpoints and credentials.

Expected Behavior

One of the following would resolve the issue:

  1. Allow Azure OpenAI node to use a fully custom base endpoint without requiring instance/resource name

  2. Support Azure OpenAI behind gateways/proxies

  3. Provide clearer validation or documentation that Azure OpenAI node does not support custom domains

Workaround

  • Expose an OpenAI-compatible endpoint (/v1/chat/completions) at the gateway

  • Use OpenAI Credential + OpenAI Chat Model instead of Azure OpenAI node

Summary

  • HTTP Request → works

  • Azure OpenAI node → fails

  • Root cause appears to be hard dependency on Azure resource name–based URL construction

  • This blocks AI Agent usage in gateway/proxy architectures

    Information on your n8n setup

  • **n8n version:**Version 1.117.3

  • Database (default: SQLite): postgreSQ:

  • n8n EXECUTIONS_PROCESS setting (default: own, main):

  • Running n8n via (Docker, npm, n8n cloud, desktop app): docker

  • Operating system:

2

Thank you to all the experts for your help and explanations. Much appreciated.