dataverse-mcp 1.0.0

A read-only MCP server for querying Microsoft Dataverse environments during development

Dataverse MCP Server

An MCP server for interacting with Microsoft Dataverse environments. Built with FastMCP and the official PowerPlatform-Dataverse-Client Python SDK.

Features

• Solution inspection — list solutions, get solution details, browse solution components
• Table querying — flexible OData-style queries against any Dataverse table
• Schema exploration — list tables, inspect table metadata (primary key, name attribute)
• Environment discovery — list Power Platform environments available to the authenticated user
• Multi-environment targeting — one MCP server config can query any Dataverse org the caller specifies
• Agent-friendly — rich tool descriptions designed for AI agent discoverability
• Secure — Pydantic v2 input validation, GUID format enforcement, OData injection prevention

Prerequisites

• uv — install from docs.astral.sh/uv
• Access to a Microsoft Dataverse environment
• Azure CLI (az login) or a registered app for authentication

Installation

You can run this server either from PyPI with uvx or directly from a local checkout.

Option 1: Run from PyPI

uvx dataverse-mcp

uvx downloads and runs the package in an isolated environment.

Option 2: Run from a local checkout

git clone https://github.com/ryanmichaeljames/dataverse-mcp.git
cd dataverse-mcp
uv sync

This creates .venv, which is the local Python environment used by the source-based MCP configuration shown below.

Configuration

Configure the server through your MCP client. In VS Code, that means the env block on the server entry in .vscode/mcp.json or your user mcp.json. This project does not use a .env file for normal setup.

Variable	Required	Default	Description
DATAVERSE_URL	No	—	Optional fallback org URL, set only if you want requests without dataverse_url to still work
DATAVERSE_AUTH_TYPE	No	azure_cli	Auth method: interactive, client_secret, or azure_cli
AZURE_TENANT_ID	For client_secret	—	Azure AD tenant ID
AZURE_CLIENT_ID	For client_secret	—	App registration client ID
AZURE_CLIENT_SECRET	For client_secret	—	App registration client secret

Authentication Methods

• azure_cli (default) — Uses your existing az login session. Best for local development.
• interactive — Opens a browser window for interactive sign-in.
• client_secret — Uses a service principal. Requires AZURE_TENANT_ID, AZURE_CLIENT_ID, and AZURE_CLIENT_SECRET.

Usage

This server communicates over stdio and works with any MCP-compatible client.

VS Code

Add the server to your VS Code MCP configuration. Choose either the packaged uvx form or the local source form.

Run from PyPI:

{
  "servers": {
    "dataverse-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["dataverse-mcp"],
      "env": {
        "DATAVERSE_AUTH_TYPE": "azure_cli"
      }
    }
  }
}

Run from a local checkout on the same machine:

{
  "servers": {
    "dataverse-mcp-local": {
      "type": "stdio",
      "command": "C:\\path\\to\\dataverse-mcp\\.venv\\Scripts\\python.exe",
      "args": ["-m", "dataverse_mcp.server"],
      "env": {
        "PYTHONPATH": "C:\\path\\to\\dataverse-mcp\\src",
        "DATAVERSE_AUTH_TYPE": "azure_cli"
      }
    }
  }
}

The local source form does not require a build step. Code changes are picked up on the next server start.

If you want a fallback environment for requests that omit dataverse_url, add DATAVERSE_URL to that same env block. Keep it in MCP config, not a .env file.

Environment Targeting

Use a single server entry and provide dataverse_url on each tool call to target the Dataverse environment explicitly. Example tool input:

{
  "dataverse_url": "https://yourorg.crm.dynamics.com",
  "table_name": "account",
  "select": ["name", "accountid"],
  "top": 10
}

If you omit dataverse_url, the server falls back to DATAVERSE_URL when that value is configured in your MCP server env. That fallback is kept for backward compatibility only; the preferred setup is explicit environment targeting on every request.

Use dataverse_list_environments first if you need to discover which Power Platform environments are available before choosing a dataverse_url.

dataverse_list_environments does not require dataverse_url and always returns the full normalized environment payload. Optional flags let you include capacity and add-on details.

Tools

Tool	Description
dataverse_list_environments	List Power Platform environments available to the authenticated user via the admin API, returning the full normalized payload
dataverse_list_solutions	List solutions with optional OData filter, select, and top
dataverse_get_solution	Get a single solution by unique name or GUID
dataverse_list_solution_components	List components in a solution with optional type filter
dataverse_query_table	Query records from any table with filter, select, orderby, expand, top
dataverse_get_record	Get a single record by table name and GUID
dataverse_list_tables	List available tables/entities with optional filter
dataverse_get_table_metadata	Get schema details for a specific table

Project Structure

src/dataverse_mcp/
├── __init__.py          # Package init
├── _app.py              # FastMCP instance (avoids circular imports)
├── server.py            # Entry point, logging setup, tool registration
├── client.py            # DataverseClient wrapper (auth, lifecycle)
├── models.py            # Pydantic v2 input models for all tools
└── tools/
    ├── __init__.py      # Tools package init
    ├── environments.py   # Power Platform environment discovery tool
    ├── solutions.py     # Solution query tools
    ├── tables.py        # Table record query tools
    └── metadata.py      # Table/column metadata tools

Development

# Install dependencies into .venv
uv sync

# Run the MCP inspector for testing
uv run mcp dev src/dataverse_mcp/server.py

# Run the server directly from source
uv run python -m dataverse_mcp.server

# Compile check touched modules
uv run python -m py_compile src/dataverse_mcp/server.py

If you run the server from a local checkout in VS Code, restart the MCP server after code changes so the new Python source is loaded.

License

MIT