A new Microsoft open-source plugin supplies top AI coding platforms with the knowledge required to build and manage agents that perform data management functions within Dataverse. The plugin works with the widely used GitHub Copilot and Anthropic’s Claude Code.
Microsoft positions the plugin — Dataverse Skills — as a way to democratize Dataverse beyond platform experts to include any developer building AI agents. Before Dataverse Skills, developers would need to juggle tools and documentation to create their required Dataverse functions; this manual approach is being replaced by a natural-language prompt that the agent uses to handle Dataverse orchestration, schemas, data, and queries.
Tapping Skills
The Dataverse Skills plugin teaches coding agents how to build and manage Dataverse solutions; it includes three skills that developers need to manage Dataverse development: connect, build, and operate.
With the plugin, a developer or user never needs to invoke skills directly; instead, they describe their intent and the agent autonomously determines the skills that are needed, loads them in the correct order, and also identifies the tools to be used to execute the required functions.
Here’s detail on how each of the three supported skills operate in this context:
Connect: The agent discovers the organization’s data environments, authenticates requests, registers the Dataverse Model Context Protocol (MCP) server — which enables connections from agents to data sources — and initializes a project structure.
Build: The agent creates tables, columns, lookups, many-to-many relationships, forms, and views. It knows the right tool for each job. For example, it can invoke MCP for quick reads, Python Software Development Kit (SDK) for bulk operations, and the Web API as needed — and adds every component to the solution automatically.
Operate: The agent loads data, runs analytical queries, bulk-imports data from spreadsheet files, and profiles data quality using the Dataverse Python SDK.
To get started with Dataverse Skills plugin, a developer initiates a Dataverse CLI terminal session, installs the plugin, and creates a prompt specifying required tables, lookups, and other details, then asks that a solution be built. In response, the agent will discover the data environment, configure MCP, build required tables, generate a script to load sample data, and conduct queries across tables to answer the underlying business question.
What the developer doesn’t need to do: reference skills names or tools or perform context-switching between docs, command line, or scripts. The developer also doesn’t need to know or describe details of orchestration requirements.
Skills are plain markdown files that contain the same knowledge and safety checks, so once downloaded, a single plugin works with both GitHub Copilot and Claude Code. Microsoft is encouraging contributions to the open-source project.
More AI Agent and Data Insights:
- Microsoft Framework Supports Diverse Models, Agent to Drive Complex Workflows
- Why Microsoft’s Support for Third-Party AI Tools Is So Important
- MCP Enablement Brings AI Automation to Dynamics 365 at Vast Scale
Tom Smith
Analyst
Cloud Wars, Agent and Copilot
Related Posts
