microsoft dataverse expands mcp to claude, cursor, and copilot with enterprise governance
A July 6, 2026 Microsoft Power Platform announcement extends the Dataverse coding-agent plugin to Claude Code, Cursor, and GitHub Copilot — joining Visual Studio. It comes with an enterprise governance layer: 60+ certified MCP servers, RBAC inheritance per user, and explicit approval gates on destructive writes. For teams already running Power Platform, it means live Dataverse data in whatever coding tool they prefer.
Microsoft announced on July 6, 2026 that the Dataverse coding-agent plugin — previously available in Visual Studio only — now works in Claude Code, Cursor, and GitHub Copilot. The expansion comes alongside a broader MCP governance push: more than 60 certified MCP servers, RBAC enforcement at both the server and tool level, and required human approval before agents can execute destructive writes against Dataverse data.
what the plugin does
The Dataverse coding-agent plugin gives your IDE a live connection to Dataverse tables, entities, and metadata through the MCP protocol. When you're building a Power Platform app or any system that reads or writes Dataverse, the agent can query your actual schema, pull live records, generate compliant code, and validate against the real data model — rather than working from a description of what the schema probably looks like.
The plugin routes each request through whatever tool is appropriate: the Dataverse MCP server for data operations, the Python SDK for scripting tasks, PAC CLI for environment management, or the Dataverse CLI for low-level operations. The routing is automatic — you don't configure it per-request.
governance: the part that matters for enterprise
The governance model is the more significant announcement for teams thinking about actually deploying this in production.
Per-user security inheritancemeans the plugin respects existing Dataverse RBAC. When you query Dataverse through Claude or Cursor, you see exactly what your Dataverse user account is authorized to see — no more, no less. The agent cannot access tables or records you couldn't access directly. This is the obvious-in-hindsight design that not every agentic integration gets right: the agent operates on your behalf, so it operates within your permissions.
Explicit approval on destructive writesmeans create, update, and delete operations require a human approval gate before they execute. The agent can propose and stage a destructive operation, but it doesn't fire until you confirm. For production Dataverse environments with real customer data, this is a meaningful safeguard against agent mistakes that would otherwise require a restore.
DLP policy enforcement is applied at both the MCP server and the individual tool level. Organizations that have existing Data Loss Prevention policies configured in Power Platform have them respected automatically — no additional configuration required to bring the coding agent within the same compliance boundary.
the 60+ mcp server catalog and certification
Alongside the Dataverse plugin expansion, Microsoft announced more than 60 ready MCP servers covering its broader Azure and M365 surface. Importantly, they introduced a certification program: partner-built MCP servers can be submitted for review and, if approved, appear in a trusted catalog. Organizations can also bring internal MCP servers under central governance — registering, auditing, and managing their own MCP tooling through the same admin surface as certified partners.
The practical implication: if your organization builds an internal MCP server that exposes, say, your HR system or internal knowledge base, you can bring it under the same governance framework as the Microsoft-certified servers. Admins get a single pane of glass over all MCP servers employees might use with their AI coding tools.
what this means if you're not a power platform shop
The honest answer: not much yet. If you're not already using Dataverse, expanding MCP support to Claude and Cursor doesn't give you a reason to start. The feature serves existing Power Platform customers who were limited to Visual Studio or manual data export workflows.
But the broader pattern is worth watching. Microsoft is systematically extending MCP governance across its entire enterprise data surface — Dataverse is the first large dataset to get this treatment, and it won't be the last. Organizations already building agentic workflows on top of Microsoft's stack are getting a governance layer that most competitors' MCP ecosystems haven't shipped yet. That gap will matter more as enterprise MCP adoption scales.