Azure OpenAI and Azure API Management

2501 agents can run on your own Azure OpenAI deployments, with model traffic routed through Azure API Management if that is how your organisation reaches them. AI use then stays inside the policy, rate limits and logging you already apply to Azure.

In many enterprises the question is not which model is best, but which model is approved. The platform has to work with that answer rather than around it.

What 2501 does

Azure OpenAI is registered in 2501 as a model provider, pointed at your endpoint or at your API Management gateway, and you choose which of your deployments the agents use. Every agent works with two models: a main engine that does the work and a secondary engine that reviews every command before it runs. Either can be an Azure OpenAI deployment. A deployment that accepts images can also read ticket attachments, and Azure providers can authenticate with Microsoft Entra ID instead of an API key.

You stay in control

• Only models you approved: agents can use only enabled models in your catalog, and disabling a provider makes all of its models unavailable at once.
• No keys in configuration files: a key lives in the credential store and can be rotated in the UI, or Entra ID can replace it.
• The model is not the guardrail: whichever model you choose, commands still pass the blacklist, which does not depend on an LLM judgment.
• You choose the level, per specialty: investigate only, act after a person confirms, or act on its own within your rules.

Frequently asked

Can 2501 use Azure OpenAI through API Management? Yes. Point the provider at your API Management gateway and model traffic goes through it.

Can 2501 also work on our Azure infrastructure? Yes. The Azure Central specialty handles maintenance, incident investigation and remediation across Azure VMs and associated services.

Related

• Azure Central
• MCP servers as agent tools
• Guardrails: what an agent may never touch
• Read-only AI investigation