Microsoft Foundry Adds Runtime, Tooling, and Governance for Production Agents
ai agentic-ai ai-agents openclaw hermes azure-ai-foundry
trending_up Trend: ai

Microsoft Foundry Adds Runtime, Tooling, and Governance for Production Agents

calendar_month June 14, 2026 update Updated: July 26, 2026

🔄 Update — 26 July 2026: Enterprise AI Agent Infrastructure with Azure Container Apps, Foundry Agent Service & Copilot Studio

Microsoft released new architectural blueprints and updates for deploying enterprise AI agents. Key developments include integrations across Microsoft Foundry Agent Service, Azure Container Apps, Copilot Studio, real-time threat protection, and remote Model Context Protocol (MCP) servers.

What’s new?

  • Serverless & IaC Agent Infrastructure: Deployment blueprints for agent runtimes running on Azure Container Apps using Terraform IaC and Dapr pub/sub messaging without hardcoded API keys.
  • Remote MCP Streaming & Copilot Studio: Enhanced Model Context Protocol (MCP) integration allowing Copilot Studio agents to connect to remote MCP servers with streaming support in Azure AI Foundry.
  • Real-Time Threat Protection: Integrated security frameworks providing real-time threat protection for enterprise agents executing autonomous API calls.

Why this adds to the article

These architectural additions bridge Microsoft Foundry’s governance and Agent Service control plane with concrete, serverless enterprise infrastructure (Azure Container Apps, Dapr, Remote MCP) and real-time security controls required for production deployments.

🔄 Update — 18 June 2026: Microsoft Introduces Agent Service, Routines, and Agent Optimizer to Microsoft Foundry

At Build 2026, Microsoft announced a major update to Microsoft Foundry to enhance the development, deployment, and observability of autonomous AI agents. Key additions include the “Agent Service” for hosting long-running agents, “Routines” for event-driven workflows, and the “Agent Optimizer” for closed-loop error-to-prompt optimization.

What’s new?

  • Agent Service & Routines: A managed hosting runtime for long-running agents and trigger-based, repeatable processes (Routines) to run asynchronous operations.
  • Procedural Memory & VS Code Toolkit: Developer tools and memory enhancements that allow agents to retain procedural knowledge across execution sessions.
  • Agent Optimizer: A closed-loop observability feature that automatically converts production agent execution failures into ranked, reviewable prompt adjustments.

Why this adds to the article

These updates build directly onto Microsoft Foundry’s runtime safety and governance framework by adding the necessary execution infrastructure (Agent Service) and feedback loops (Agent Optimizer) needed to run and monitor enterprise-scale autonomous agents in production.

Summary

At Build 2026, Microsoft announced substantial updates to Microsoft Foundry, equipping it with runtime security, stateful session management, open protocol tooling (MCP), and enterprise-grade governance. The platform now supports long-running agents like OpenClaw and Hermes, enabling scheduled routines for asynchronous workflows.

Details

Microsoft has rolled out stateful session sandboxes, governance templates via Azure, and deep integration of open protocols like Model Context Protocol (MCP) in Microsoft Foundry Agent Service. This setup enables executing long-running agents on recurring schedules, providing a standardized hosting and security layer for autonomous agents inside the enterprise.