Agents-as-a-Service Rush
Companies are moving beyond writing copilots to AI agents that complete end-to-end tasks for enterprise workflows, and vendors are racing to make those agents safe and auditable. Microsoft released Azure MCP Server 2.0 with remote self‑hosting and hundreds of tools for enterprise AI automation, while analysts describe the shift as one from model choice to platform control and authorization standards are being drafted to govern agent access to resources. Shopify has also started opening backend systems to agent integrations such as Claude, showing the trend is already touching large e‑commerce stacks. (moneycontrol.com) (c-sharpcorner.com) (dasroot.net) (x.com)
Enterprise software companies are selling artificial intelligence agents that do the work, not just copilots that suggest it. Microsoft and Shopify both shipped new agent plumbing this week for cloud and commerce workflows. (moneycontrol.com) (devblogs.microsoft.com) (shopify.dev) Microsoft said on April 10 that Azure Model Context Protocol Server 2.0 is now in stable release with self-hosted remote deployment over Hypertext Transfer Protocol Secure, aimed at team and enterprise use. The company said the server exposes 276 tools across 57 Azure services for tasks including provisioning, deployment, monitoring, and diagnostics. (devblogs.microsoft.com) (learn.microsoft.com) Model Context Protocol is the connector layer that lets an artificial intelligence system call outside tools the way a travel agent calls an airline system. Microsoft’s release focuses on remote hosting, authentication, and centralized policy controls so those tool calls can run inside enterprise guardrails instead of on one developer laptop. (learn.microsoft.com) (devblogs.microsoft.com) That security work is moving into standards. The Model Context Protocol specification says authorization implementations must use OAuth 2.1, and recent work around resource indicators is meant to keep an agent’s token scoped to the specific system it is allowed to touch. (modelcontextprotocol.io) (dasroot.net) The commercial pitch has shifted with the technology. Moneycontrol reported that analysts now see enterprise buyers moving from choosing the best model to choosing the platform that controls access, approvals, audit trails, and billing for agent work. (moneycontrol.com) Shopify pushed the same idea into e-commerce on April 9 with its Shopify AI Toolkit. Shopify said the toolkit connects Claude Code, Cursor, Gemini Command Line Interface, Visual Studio Code, and Codex Command Line Interface to Shopify documentation, application programming interface schemas, code validation, and store management through command-line execution. (shopify.dev 1) (shopify.dev 2) Shopify’s public documentation says developers can install the toolkit as an auto-updating plugin or use agent skills or a development Model Context Protocol server instead. Its GitHub repository shows skill folders for admin execution, customer data, payments apps, point-of-sale user interface, Liquid templates, Hydrogen, and Storefront GraphQL. (shopify.dev) (github.com 1) (github.com 2) The promise is fewer handoffs inside software. Moneycontrol described agents that can execute end-to-end business tasks, while Deloitte wrote in late 2025 that software-as-a-service products were evolving toward federated workflow services that are more autonomous and adaptive. (moneycontrol.com) (deloitte.com) The constraint is trust. Microsoft’s own engineering blog said the Model Context Protocol specification had been changing rapidly to address security concerns, and the current authorization spec separates the tool server from the authorization server so credentials can be scoped, rotated, and audited. (devblogs.microsoft.com) (modelcontextprotocol.io) The race now is less about adding one more chatbot button to a dashboard. It is about which vendor becomes the control plane for agents that can read systems, take actions, and leave a paper trail when they do. (moneycontrol.com) (devblogs.microsoft.com)
