---
name: Azure AI Foundry Setup
description: Azure credentials path + endpoints for Icedout Solutions LLP's Azure AI Foundry resource
type: reference
originSessionId: 38a0dcdb-5376-45aa-a4f3-1527accc3abe
---
## Credentials location
- **API key file:** `~/.azure_key` (chmod 600) — NEVER paste key contents into memory or chat
- Use pattern: `AZURE_KEY=$(cat ~/.azure_key)` in scripts
- If compromised: https://portal.azure.com → arnav-resource → Keys and Endpoint → regenerate → overwrite `~/.azure_key`

## Resource details
- **Resource name:** `arnav-resource`
- **Project name:** `arnav` (within Azure AI Foundry)
- **Region:** (TBD — confirm from portal)
- **Subscription:** Microsoft for Startups Founders Hub ($1,000 credits starting tier, acquired April 2026)

## Endpoints

### Azure AI Foundry Project endpoint
```
https://arnav-resource.services.ai.azure.com/api/projects/arnav
```
Use this for:
- Agents, threads, assistants API
- Evaluations
- Model inference via unified Foundry API
- Azure AI Agents SDK

### Azure OpenAI v1 endpoint
```
https://arnav-resource.openai.azure.com/openai/v1
```
Use this for:
- Direct Azure OpenAI Service access (GPT-4o, GPT-5 when available, DALL-E, etc.)
- OpenAI-compatible SDK calls
- Chat completions, embeddings, images

## Authentication
Include in request headers:
```
api-key: $(cat ~/.azure_key)
```
Or use Azure AD / Entra ID auth if preferred.

## SDK setup examples

### OpenAI-compatible client (Python)
```python
from openai import OpenAI

client = OpenAI(
    base_url="https://arnav-resource.openai.azure.com/openai/v1",
    api_key=open("/home/mail2arnav99/.azure_key").read().strip(),
)
```

### Azure AI Foundry project (Python)
```python
from azure.ai.projects import AIProjectClient
from azure.core.credentials import AzureKeyCredential

project = AIProjectClient(
    endpoint="https://arnav-resource.services.ai.azure.com/api/projects/arnav",
    credential=AzureKeyCredential(open("/home/mail2arnav99/.azure_key").read().strip()),
)
```

## Usage tips
- **Budget alerts:** Set up at portal.azure.com → Cost Management → Budgets (avoid Gemini-style surprise spend)
- **Deployment names:** Each model needs a deployment name set in Azure AI Foundry before you can call it (e.g., `gpt-4o-deployment`, not just `gpt-4o`)
- **Rate limits:** New resources often start with conservative quotas — request increases via Azure portal

## Context: what Microsoft for Startups unlocked
- $1,000 Azure credits (12-month validity)
- GitHub Enterprise seat
- LinkedIn Premium
- Figma / Notion / HubSpot / MongoDB / Stripe credits
- Path to upgrade: $5K (Grow tier) → $25K (Scale) → $150K (AI startup tier)
- Upgrade requires active Azure usage + product validation / funding