Microsoft offers Foundry Agent Service as a managed platform on Azure for building, deploying, and running AI agents in production environments, with built-in governance and monitoring capabilities and integration with enterprise systems. The service allows teams to use the framework and model they choose, then deploy the agent with a single command within a managed, session-isolated runtime environment.
The service targets teams that want to move from local code or prototypes to running agents at enterprise scale without building the infrastructure required to manage these environments. The service supports automatic scaling and does not charge for idle periods for hosted agents, according to Microsoft’s product page.
Running Agents and Connecting Them to Tools and Data
Foundry Agent Service enables the orchestration of multi-agent workflows and connects agents to enterprise knowledge through Foundry IQ, including SharePoint, Fabric, and Bing. It also provides connections to more than 1,400 tools that support the MCP protocol, along with managed short-term and long-term memory and procedural memory that enable context retention without the need to create and manage an external database.
The service is based on open standards such as MCP, A2A, and OpenAPI, helping teams reuse integrations and avoid being locked into a single protocol. It also includes Toolbox for grouping reusable tools and exposing them through an MCP-compatible endpoint, with centralized management of authentication, settings, and versioned updates.
Identity, Security, and Monitoring
The service supports assigning each agent an independent identity through Entra Agent ID, with capabilities for access control, identity lifecycle management, and flexible authentication for accessing tools. Microsoft provides private networking options, including the use of a customer-owned virtual network, with the ability to run agents without public data egress.
Monitoring capabilities include tools for tracking interactions, linking execution traces, evaluating agent quality, and applying controls and policies. These functions are intended to help teams understand agent behavior, troubleshoot issues, and enforce safeguards during operation.
Deployment and Integration with the Work Environment
Agents can be deployed to Microsoft Teams and Microsoft 365 Copilot with one click, with their visibility managed through governance covering the entire tenant. The service also supports integration with Voice Live to build real-time voice interaction experiences based on speech-to-speech conversion.
Microsoft explains that hosted agents provide a managed runtime environment for agent code in Azure and can be used with frameworks such as Microsoft Agent Framework and LangGraph. Developers can also choose the model best suited to the task and benefit from integration with Azure OpenAI in Foundry Models, Microsoft Foundry, Azure AI Search, and Azure API Management.
Pricing and Use Cases
Foundry Agent Service pricing is based on the individual models and tools that the user accesses and uses. The use cases Microsoft cites include automating customer support through integration with customer relationship management systems, automating routine tasks and workflows such as scheduling meetings, sending reminders, and creating reports, as well as providing insights and recommendations based on real-time data analysis.
Microsoft says Foundry is used by more than 80,000 organizations and digital startups, including 80% of Fortune 500 companies, and also points to more than 3 billion enterprise search queries per day. The service provides documentation, tutorials, use cases, and resources to help teams build, deploy, and manage agents.