What It Means When the Agent Runtime Becomes Free
Microsoft’s move to make the agent runtime free means the execution loop that runs AI agents is now treated as shared infrastructure, while value and revenue shift to higher layers that control identity, policy, observability, and integration across enterprise systems. At Build, Microsoft shipped Scout, its first always-on work agent, on OpenClaw, an open-source runtime originally hacked together over a weekend by an independent developer. Microsoft is contributing its enterprise policy controls back to OpenClaw, underscoring that the runtime itself is no longer where it expects to earn money. Instead, Scout sits inside a larger managed environment that connects to Microsoft 365 data, Model Context Protocol tools, and Windows Execution Containers. The strategy echoes Android: make the base free and focus on the Microsoft control plane, managed identity, and surrounding services that enterprises will pay to standardize on at scale.
From Runtime to Control Plane: Where Microsoft Plans to Earn
By adopting OpenClaw rather than building its own loop, Microsoft signals that the agent runtime free layer is a commodity and the control plane is the product. Enterprises care less about how an invoice-reading loop runs and more about who the agent is, what it is allowed to touch, and how its actions are audited. Scout agents run under governed Entra identities instead of shared service accounts, closing the “agentic identity crisis” where autonomous agents once acted on borrowed credentials. Policies restrict what each agent can do, while logs and observability give auditors a clear trace of every action. According to The New Stack, Microsoft is positioning itself for “every layer except the runtime,” a stack that includes device-level containment with Microsoft Execution Containers and multi-agent governance that can extend to third-party runtimes like Nvidia’s OpenShell and Nous Research’s Hermes Agent.
Logic Apps Automation: SaaS Orchestration for Enterprise AI Agents
Azure Logic Apps Automation turns enterprise AI orchestration into a managed SaaS experience, offered through auto.azure.com. Instead of assembling compute, connectors, model endpoints, and knowledge stores one by one, teams sign in to a ready-made environment where these components are pre-wired and governed. The new SKU targets the gap between “every team has an AI agent demo” and the far smaller number that reach production. Logic Apps Automation gives each project an isolated compute boundary, with identity, RBAC, audit logging, networking controls, VNET integration, and private endpoints on by default. Agents integrate in three ways: as tools inside an agent loop, as Microsoft Foundry agents called directly from the workflow canvas, or via a managed sandbox for popular agent harnesses. This moves the business model away from charging for runtimes and toward selling a full-stack, production-ready orchestration layer for AI workflows and agents.
Microsoft Foundry: Turning Experiments into Production Agents
Microsoft Foundry is the company’s AI app and agent factory, and its new features clarify how Microsoft Foundry agents fit into the revenue picture. Foundry now offers runtime, tools, memory, grounding, models, observability, and governance in one platform that moves agents from experiments into production systems. Hosted agents run in sandboxed sessions with state and filesystem access, and the same runtime supports long-running agents like OpenClaw and Hermes, plus scheduled routines for tasks such as overnight ticket triage. Toolboxes provide a managed endpoint for tools, skills, Model Context Protocol clients, and enterprise data integrations, so agents discover capabilities at runtime instead of hard-wiring every tool. According to InfoQ, Foundry also supports direct publishing into Microsoft Teams and Microsoft 365 Copilot, so the control plane extends all the way to the user interface, with shared policy and observability applied automatically across every agent deployment.
The New Revenue Model for Enterprise AI Orchestration
Taken together, Scout on OpenClaw, Logic Apps Automation, and Foundry outline Microsoft’s new revenue model for enterprise AI agents. The agent runtime free layer becomes a shared substrate that multiple vendors can adopt, while Microsoft charges for the Microsoft control plane and managed services wrapped around it. Logic Apps Automation bundles workflows, AI agents, knowledge services, and model access into a single governed SaaS, aimed at business teams that need production-grade outcomes without specialist integration skills. Foundry supplies the lifecycle: building, grounding, observing, governing, and distributing agents at scale. In this model, the money is in compliance, identity, networking, data access, and day-two operations rather than in the core agent loop. As more runtimes like OpenClaw and OpenShell thrive as open bases, enterprises will pay for the orchestration, not the kernel, and Microsoft intends to own that layer end to end.
