Microsoft published its June 2026 Power Platform feature update this week, and the theme is hard to miss. The platform is no longer just adding maker features. It is building the governance and learning infrastructure that AI agents at enterprise scale actually require. Here is what matters from this round, with a focus on the items that change how you administer and govern a tenant.
Advanced connector policies are now generally available
The headline for governance teams is that advanced connector policies (ACP) have reached general availability. This is the shift from governing which connectors are allowed to governing which actions and which MCP servers inside those connectors AI tools are permitted to use. If your tenant grew from a few dozen environments to several thousand on the back of Copilot and agent projects, blanket connector blocks are too blunt an instrument. ACP gives you action level and MCP level control, which is the granularity the agent era demands.
If you have been following the blog, you know ACP went GA on June 4. Seeing it carried into the monthly roundup confirms Microsoft is positioning it as a core governance primitive, not a niche feature.
Inventory now shows connectors and connector operations
This is the update I would push to the top of your reading list. In public preview since June 2, Power Platform inventory now captures the connectors and connector operations used by every canvas app, model-driven app, cloud flow, agent flow, Microsoft 365 agent flow, and agent built in either Copilot Studio or Microsoft 365 Copilot Agent Builder. For flows it also records the trigger connector and trigger operation. For agents it captures richer metadata about how each connector is used as a tool or as a knowledge source.
The practical value here is enormous. Inventory used to tell you what existed in your tenant. Now it tells you what each resource talks to. That means you can pinpoint every resource affected by a connector deprecation, a security issue, or a licensing tier change in seconds. You can see which connectors dominate adoption across the tenant. And critically, you can ground your ACP decisions in real usage data before you tighten or block anything, so you know exactly which apps, flows, and agents you are about to impact.
A new connectors column appears across the inventory grids in the Power Platform admin center, including the unified manage and inventory page and the resource specific views under Copilot Studio, Power Apps, and Power Automate. The same data is queryable through the Power Platform for Admins V2 connector, the Power Platform inventory API, and Azure Resource Graph, which turns tenant wide connector analysis into a single query rather than a manual project. No opt in is required. The data flows into inventory automatically.
Power Apps MCP server gets closed-loop learning
For anyone running agents at scale, this one is worth understanding. Microsoft introduced closed-loop learning for agents connected to the Power Apps MCP server, starting with the data entry tool. Every correction a user makes through the agent feed now persists as structured memory. On future runs the agent retrieves that memory and applies it, and over time those corrections consolidate into organization wide patterns.
The part that stands out is that the feedback loop runs automatically in production. There are no data pipelines to build and nothing to configure. This addresses a real ceiling enterprises hit, where teaching an agent how the organization actually works meant either feeding in documents and custom instructions endlessly or standing up a data science team to manage the training and evaluation cycle. From a governance angle, this also means agent behavior will drift based on user corrections, so it is worth tracking how that consolidated memory is reviewed and audited over time.
Power Automate desktop: version comparison and a run Power App action
Two solid additions for the RPA crowd. First, Power Automate for desktop now includes side by side version comparison as part of built-in version control. You can compare two versions of a desktop flow across subflows, actions, variables, UI elements, and images, and search within the comparison view to find specific changes. Versions are stored in Dataverse and retained for up to 12 months, which gives you a properly governed change history for desktop automations rather than the informal versioning many teams rely on today.
Second, in public preview, a new run Power App action lets a desktop flow open a Power App directly and establish a native communication channel between the two. You can pass inputs from the flow into the app, capture outputs back, and trigger callable subflows from app events. This is available in Power Automate for desktop version 2.68 or later, and it opens up attended automation scenarios like guided forms and app based front ends without the UI based workarounds people have been forced into.
Release planner code app
The Power CAT team shipped a sample release planner code app built with Power Apps code apps, React, and custom connectors. It lets you search, filter, and track upcoming features across Power Platform, Dynamics 365, and Microsoft Copilot by product, release wave, cloud availability, and status, with card, calendar, and timeline views. Beyond its day to day usefulness for staying ahead of the roadmap, it is a good reference for how to build enterprise grade experiences with code apps while integrating external APIs.
The takeaway
If there is a through line in this update, it is that Microsoft is building the connective tissue for governing AI at scale. ACP gives you the controls, inventory connector visibility gives you the data to use those controls intelligently, and closed-loop learning shows where agent behavior is headed. Admins who pair the new inventory connector data with ACP will be in a far stronger position than those still governing at the connector level alone. That is the work worth prioritizing this month.