Azure OpenAI Assistants Azure AI Foundry Agent Service
Legacy names Azure OpenAI Assistants, Azure AI Agent Service
Managed stateful agent runtime for building, deploying, and scaling AI agents with persistent threads, toolboxes, file search, code interpreter, MCP servers, and multi-model orchestration.
Naming timeline
- Azure OpenAI Assistants
Launched in preview within Azure OpenAI Service providing stateful threads, code interpreter, and function/file tools.
- Azure AI Agent Service
Announced as an enterprise agent orchestration service within Azure AI Foundry uniting Azure OpenAI Assistants capabilities with broader model and enterprise tool integrations.
- Azure AI Foundry Agent ServiceCurrent name
Reached general availability as Azure AI Foundry Agent Service (also styled as Foundry Agent Service in Microsoft Foundry), superseding the classic Azure OpenAI Assistants API preview.
What is Azure OpenAI Assistants called now?
Azure OpenAI Assistants is now called Azure AI Foundry Agent Service, with the latest change on 19 May 2025 (Microsoft Build 2025 GA). Earlier names covered by this entry also include Azure AI Agent Service. Managed stateful agent runtime for building, deploying, and scaling AI agents with persistent threads, toolboxes, file search, code interpreter, MCP servers, and multi-model orchestration. Only the customer-facing name changed; the underlying technical identifiers remain backward compatible.
Also known as
- Foundry Agent Service
- Microsoft Foundry Agent Service
- Azure OpenAI Assistants API
Technical identifiers
Technical identifiers unchanged. Only the customer-facing product name changed; the APIs, CLI commands, IAM roles and Terraform resources below keep their legacy identifiers for backward compatibility.
| ARM Resource Providers (Microsoft.*) | Microsoft.CognitiveServices/accounts Microsoft.CognitiveServices/accounts/projects Microsoft.MachineLearningServices/workspaces |
|---|---|
| Azure CLI (az) / PowerShell | az cognitiveservices account az ml workspace |
| Entra ID / Azure RBAC Roles | Azure AI Developer Azure AI User Cognitive Services OpenAI User Cognitive Services OpenAI Contributor |
| Terraform Resources (azurerm_*) | azurerm_ai_foundry azurerm_ai_foundry_project azurerm_cognitive_account |
Agent runtimes operate on Microsoft.CognitiveServices/accounts (and Microsoft.MachineLearningServices/workspaces for hub projects) via the azure-ai-projects and azure-ai-agents SDKs, while legacy Azure OpenAI Assistants preview endpoints transition to Foundry Agent Service.
Notes & disambiguation
- Originally introduced in public preview in February 2024 as the Azure OpenAI Assistants API (Preview) for stateful thread management, code interpreter, and file search on Azure OpenAI models, the capability was expanded into Azure AI Agent Service at Microsoft Ignite (November 19, 2024) and reached general availability as Azure AI Foundry Agent Service (also styled as Foundry Agent Service / Microsoft Foundry Agent Service) at Microsoft Build on May 19, 2025.
- The standalone classic Azure OpenAI Assistants API (Preview) is retired in favor of the generally available Foundry Agent Service, which adds multi-model support across the Foundry model catalog (GPT, Llama, DeepSeek, Mistral), managed Hosted Agents, Voice Live speech-to-speech agents, Model Context Protocol (MCP) toolboxes, and Agent-to-Agent (A2A) interoperability.
- Underlying infrastructure provisions through Microsoft.CognitiveServices/accounts (Foundry resources and projects) and Microsoft.MachineLearningServices/workspaces (hub-based Foundry projects), managed via the Azure.AI.Agents / azure-ai-projects SDKs, az cognitiveservices / az ml CLI commands, and Azure AI Developer RBAC roles.
Sources & official documentation
Catalog last updated . Spotted an error? Suggest an edit