Bring your own Azure OpenAI key
Connect your own Azure OpenAI Service resource to Meridian. Your prompts, completions, and embeddings stay inside your Azure tenant boundary — Meridian never sees your API key or request payloads.
Prerequisites
- 1An active Azure subscription with Azure OpenAI Service approved. Request access at aka.ms/oai/access if you have not already.
- 2A deployed model in your Azure OpenAI resource (GPT-4o, GPT-4o-mini, or GPT-4-Turbo). Note the deployment name — you will need it during setup.
- 3Meridian Team plan or higher. BYOK is not available on the Free or Pro tiers.
Step 1 — Gather your Azure credentials
From the Azure Portal, navigate to your OpenAI resource and copy the following values:
| Field | Where to find it | Example |
|---|---|---|
| Endpoint | Keys and Endpoint blade | https://my-resource.openai.azure.com/ |
| API key | Keys and Endpoint blade — Key 1 or Key 2 | a1b2c3d4... |
| Deployment name | Model deployments blade | gpt-4o |
Step 2 — Add your key in Meridian
- 1Open Settings → AI Providers in the Meridian dashboard.
- 2Click Add Azure OpenAI and paste your endpoint, API key, and deployment name.
- 3Meridian validates the connection by sending a single zero-token test completion. Your key is encrypted at rest with AES-256-GCM before storage.
- 4Select your Azure deployment as the default provider for new threads. You can still switch per-thread in the model picker.
Data residency and compliance
Your tenant, your data
All inference requests route directly from the Meridian client to your Azure OpenAI endpoint. Meridian servers never proxy, inspect, or log your prompt contents or completions.
Azure compliance scope
Your existing Azure compliance certifications (SOC 2, HIPAA, ISO 27001) extend to your OpenAI usage. Meridian does not alter your Azure data processing agreement.
Key encryption
Your Azure API key is encrypted client-side with a hardware-bound secret before it ever leaves your machine. Even Meridian cannot decrypt it.
Audit logging
Azure OpenAI provides detailed usage logs inside your subscription. Monitor token consumption, rate limits, and content filtering independently of Meridian.
Rate limits and quotas
Meridian respects the rate limits configured on your Azure OpenAI deployment. We recommend the following minimum quotas for a smooth Team experience:
| Model | Tokens per minute | Requests per minute |
|---|---|---|
| GPT-4o | 300K | 60 |
| GPT-4o-mini | 2M | 120 |
| GPT-4-Turbo | 80K | 30 |
Quotas can be adjusted in the Azure Portal under your OpenAI resource → Quotas blade. Increases typically require a support ticket.
Troubleshooting
Connection test fails with 401
Verify your endpoint URL ends with a trailing slash and your API key has not been regenerated. Azure keys do not expire, but manual rotation invalidates the old key immediately.
Deployment not found (404)
Confirm the deployment name matches exactly — it is case-sensitive. The deployment must be in the same region as your endpoint.
Content filter blocking requests
Azure OpenAI applies content safety filters by default. If legitimate requests are blocked, adjust your content filter policy in Azure AI Foundry or contact Azure support for modified abuse monitoring.
Need help with Azure setup?
Our solutions engineers can walk you through the Azure OpenAI provisioning process and help configure optimal quotas for your team.
Contact solutions engineering