Microsoft said in a November 2, 2026, post that Dataverse was available through Copilot Cowork, additional coding agents, and n8n. The routes serve different tasks: Cowork helps build apps through an approval-driven conversation, coding agents use shared Dataverse skills, and n8n connects through a built-in workflow node.

Microsoft also said it expanded its coding-agent plugin to Claude, Cursor, and GitHub Copilot in July 2026. Its November post lists those tools alongside other coding-agent integrations, including OpenAI Codex and ChatGPT, Gemini CLI, and Google Antigravity.

How the four connection paths differ

Integration routeHow it connects to DataverseDescribed workflow
Copilot CoworkA Dataverse pluginDesigns tables and relationships, can reuse tables such as Contact, seeds sample data, and builds and publishes an app in Microsoft Copilot Managed Runtime, which Microsoft labeled preview.
Coding agentsMicrosoft’s shared, open-source Dataverse skillsProvides skills for working with records, data models, solutions, security, environments, and Finance and Operations X++ development.
n8nA built-in Microsoft Dataverse nodeConnects Dataverse record operations to automation workflows.
Copilot Studio MCP labAn OAuth connection to the Dataverse MCP ServerA separate agent setup for querying and updating Dataverse records.

In Cowork, Microsoft says each proposed change is presented for approval and actions run under the user’s permissions. For coding agents, Microsoft provides the skills through a shared repository. For other agents that can read local skills, its installation command is npx skills add microsoft/Dataverse-skills -s "*".

These are different integration approaches: the coding-agent skills cover a range of Dataverse development and administration tasks, while the n8n node is built for connecting record operations to workflow automation. Microsoft names the Dataverse MCP server, Dataverse CLI, Python SDK, and PAC CLI among the tools its coding-agent plugin can use.

What the n8n node can do

Microsoft’s n8n demo shows a lead-capture workflow and Dataverse node configuration

Microsoft lists six operations for the n8n node: create, update, upsert, get, get many, and delete. They apply to standard and custom tables through the Dataverse Web API. For “Get Many,” Microsoft describes column selection, OData filters, sorting, expand, and FetchXML. The OAuth2 credential supports delegated user access and app-only access. The n8n Microsoft Dataverse node documentation is the product reference for the node.

Microsoft’s demonstration uses a lead-capture workflow: it takes input from a web form or messaging channels, extracts details in an AI step, checks Dataverse for an existing lead, and creates a record when needed. It also shows node and credential configuration.

The Copilot Studio MCP lab uses a separate setup

A Microsoft-hosted Copilot Studio lab describes a distinct route through the Dataverse MCP Server. Its sequence is to create the agent before adding the MCP server. With the default end-user credentials, the first test prompts for consent. The lab describes maker-provided credentials as an option for service-principal or client-certificate connections, and Activity Map lets users inspect tool calls and their inputs and outputs.

What Microsoft says about permissions and security

For its coding-agent plugin, Microsoft says it follows least privilege, Dataverse authorizes each API call on the server, and credentials stay in the operating system’s secure credential store or in memory. Microsoft also says telemetry excludes prompts, tool arguments, and record data.