Microsoft Dataverse MCP refers to the Dataverse Model Context Protocol (MCP) Server, an integration feature within Microsoft’s Dataverse platform (part of the Power Platform).
Launched in public preview in mid-2025, it enables seamless, secure connections between large language model (LLM) applications—like AI agents and copilots—and enterprise data stored in Dataverse.
MCP is an open protocol that standardizes how AI tools access external data sources, transforming structured business data (e.g., tables, records) into interactive, queryable knowledge for AI-driven workflows.
Key Features
- Natural Language Access: Users can query data conversationally (e.g., “Show me my contacts” or “Apply a 10% discount to Widget123”) without writing code or SQL. The MCP client translates these into Dataverse operations.
- Built-in Tools: Supports operations like reading/updating tables, searching knowledge, viewing prompts, and schema management (e.g., create/update/delete tables, columns, relationships, and option sets).
- Multi-Agent Support: Facilitates multi-agent operations, allowing AI agents to reason across data, execute actions, and generate outputs while respecting Dataverse security (e.g., RBAC, MFA).
- Platform-Agnostic: Works with various MCP clients, making Dataverse data extensible beyond Microsoft tools.
Supported MCP Clients
Dataverse MCP integrates with:
- Microsoft Copilot Studio: For building custom AI agents that interact with Dataverse in real-time.
- Visual Studio Code with GitHub Copilot: Enables developers to query and manipulate data during coding sessions.
- Claude Desktop: Supports third-party AI tools for natural language data access.
- Others: Including VS Code extensions and emerging LLM apps.
Setup and Availability
To use it, enable the MCP server in your Power Platform environment and configure allowed clients via the admin center. Setup involves creating a Dataverse connection, noting the environment URL, and authenticating (e.g., via delegated users or service principals). Labs and guides are available on GitHub for hands-on testing.
As of November 2025, it’s in public preview (not recommended for production), with ongoing enhancements like a remote MCP server for simpler setups and improved security. It’s designed for innovators exploring AI-data integration in enterprise apps like Power Apps and Dynamics 365.
