Power Apps MCP Server Public Preview for Agent Automation

Introduction: Why this matters

As organizations adopt AI agents for everyday business work, the biggest hurdle is operational trust: how do you let agents act quickly without losing control, auditability, or user confidence? Microsoft’s new Power Apps MCP (Model Context Protocol) Server in public preview targets that gap by enabling agent-driven automation in Power Apps while keeping humans in the loop via a redesigned enhanced agent feed.

What’s new

Power Apps MCP Server (public preview)

The Power Apps MCP Server packages Power Apps’ existing “agentic data entry” technology into a reusable MCP tool so agents can interact with apps at enterprise scale.

Key capabilities include:

• Automate repetitive app tasks (starting with data entry): Configure an MCP tool in your agent and use invoke_data_entry to pull information from unstructured sources (e.g., a shared mailbox, SharePoint folder) and create records in a Power App.
• Human review and approval built-in: Data captured by the agent is presented for verification before records are created.
• Supervision and handoff: Agents can explicitly hand control to humans for review, assistance, or steering when confidence is low or business rules require it.
• Roadmap expansion: Microsoft is starting with data entry and plans to add more operations over time (e.g., broader create/read/update/delete capabilities).

Enhanced agent feed with native MCP integration

To support MCP-driven workflows, Microsoft has overhauled the agent feed to be more task-centric and controllable.

Highlights include:

• Granular maker control over which tasks appear in the feed and when to require human-agent handoff.
• Side-by-side comparison view for data entry tasks to speed review/approval.
• Direct navigation to specific app records for context-aware validation and follow-up.
• Agent performance metrics and insights surfaced within the app experience.

Impact on IT admins and end users

• For admins: Expect increased interest from makers and business teams in deploying agent-enabled experiences inside existing Power Apps. Governance will matter—especially around connectors, access to mailboxes/SharePoint locations, environment strategy (early release vs. standard), and audit requirements for agent actions and approvals.
• For makers/business users: This enables “agentic apps” where users can supervise and collaborate with agents directly in the app workflow instead of managing automation elsewhere. The agent feed becomes the control plane for approvals and exceptions, reducing manual rekeying while preserving accountability.

Action items / next steps

• Validate availability: The preview starts in United States early release cycle environments and rolls out to additional regions on the standard weekly deployment cadence.
• Pilot a controlled scenario: Start with a low-risk, high-volume process like email/attachment-driven data entry (claims, service requests, onboarding forms) that benefits from human review.
• Plan governance: Define who can publish tasks to the agent feed, what requires approval, and what data sources (mailboxes/SharePoint) are permitted.
• Train reviewers: Ensure business approvers understand the side-by-side review experience and how to handle exceptions.
• Provide feedback: Share learnings and supervision requirements via the Power Platform Community to influence upcoming supervision and CRUD expansion.