Guide: How to connect AI/ML API to Cursor (Azure path)

How To
1

This guide explains how to connect AI/ML API to Cursor using the Azure OpenAI-compatible flow. You’ll get a clean setup with one endpoint, support for slashes in deployment names.

:rocket: Quick Setup

Field Value
Base URL https://api.aimlapi.com
API Key Your AI/ML API key (create at aimlapi.com/app/keys)
Deployment google/gemini-2.5-pro (slashes allowed)
Alias (Model ID) gpt-4o (Bypasses the restriction and makes Cursor work with any model*)*

Do not add /v2/azure or /openai to the Base URL.

:white_check_mark: Prerequisites

  • AI/ML API key

  • Cursor IDE (latest)

  • Internet access to api.aimlapi.com

1) Configure Cursor (Azure path)

Open Cursor → Settings → Models → Azure and fill in:

Base URL

Copy

https://api.aimlapi.com

Deployment Name

Copy

google/gemini-2.5-pro

API Key Paste your AI/ML API key exactly (avoid spaces).

Click Verify to confirm.

1030×633 23.2 KB

941×1086 117 KB

678×210 22.9 KB

2) Keep the model picker clean

In Cursor’s Chat model selector, only enable:

Copy

gpt-4o

This alias (Model ID) will send traffic to your deployment (google/gemini-2.5-pro).

3) How Cursor calls AI/ML API

Example request generated by Cursor:

Copy

POST https://api.aimlapi.com/openai/deployments/google/gemini-2.5-pro/chat/completions?api-version=2024-12-01-preview
Api-Key: <YOUR_AIMLAPI_KEY>
Content-Type: application/json

{
  "messages": [
    { "role": "system", "content": "You are a helpful coding assistant." },
    { "role": "user",   "content": "Write a Python function that reverses a string." }
  ]
}

Notes:

  • Deployment Name is inserted into /deployments/<NAME>/....

  • api-version is handled by Cursor automatically.

  • Base URL stays exactly https://api.aimlapi.com.

4) Optional smoke test

Copy

curl -sS -X POST \
  "https://api.aimlapi.com/openai/deployments/google/gemini-2.5-pro/chat/completions?api-version=2024-12-01-preview" \
  -H "Api-Key: YOUR_AIMLAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role":"system","content":"You are a helpful coding assistant."},
      {"role":"user","content":"Give me a one-line Python function to merge two dicts."}
    ]
  }'

You should receive a JSON response with choices[0].message.content.

5) Common pitfalls

  • Deployment not found → Check Base URL & Deployment Name.

  • Invalid API key → Re-copy the key, ensure it’s in the Azure section.

  • Wrong model list → Toggle Azure off/on, click Verify, restart Cursor.

  • Slashes in names → Allowed in Deployment, but keep alias short (e.g. gpt-4o).

6) Tips for teams

  • Standardize the alias (Model ID) (gpt-4o) so everyone sees the same thing in Cursor.

  • Document your Base URL + Deployment in team wiki to avoid drift.

  • You can swap deployments later without changing the alias in UI.

:white_check_mark: Summary (copy/paste)

  • Base URL: https://api.aimlapi.com

  • API Key: your AI/ML API key

  • Deployment: google/gemini-2.5-pro (slashes allowed)

  • Alias (Model ID): gpt-4o

With this setup, Cursor talks to AI/ML API using the Azure flow, while you keep the UI clean and consistent

2

Interesting service! How does it handle reasoning models such as GPT-5?

In particular, setting different reasoning efforts and verbosity levels, and being able to switch between them in Cursor easily? Also, does it display the resoning summaries?

Sorry for the questions, I wanted to try for myself but I think I need a pro account to be able to try GPT-5.

(full disclamer, I’m the maintainer of GitHub - gabrii/Cursor-Azure-GPT-5: A service that allows Cursor to use Azure GPT-5 deployments.)