powerbi-mcp

A small Model Context Protocol (MCP) server for Microsoft Power BI. It lets an MCP-capable client (Claude, VS Code, etc.) discover your workspaces and datasets, read a semantic model's structure, and run DAX queries — all through the public Power BI and Microsoft Fabric REST APIs.

Authentication uses MSAL with the Windows broker (WAM): you sign in once, the token is cached locally, and the server refreshes it silently afterwards. No tokens are ever pasted by hand or stored in config.

Tools

Prerequisites

• Python 3.10+
• A Power BI account with access to at least one workspace
• Windows recommended — the interactive sign-in uses the OS broker (WAM). On macOS/Linux it falls back to the system browser.

Setup (step by step)

1. Install

git clone https://github.com/rajivdatta/powerbi-mcp.git
cd powerbi-mcp
python -m venv .venv
.venv\Scripts\activate          # Windows  (source .venv/bin/activate on macOS/Linux)
pip install -r requirements.txt

2. (Optional) Configure your tenant

All configuration is optional and via environment variables — copy .env.example to .env if you want to set any:

Finding your tenant ID: Azure Portal → Microsoft Entra ID → Overview → Tenant ID, or just use your email domain. Leaving it unset (organizations) lets any work/school account sign in.

.env is git-ignored.

3. Sign in once

.venv\Scripts\activate
python server.py --login

This opens an interactive sign-in (Windows broker popup or browser) and caches a refresh token at %LOCALAPPDATA%\powerbi-mcp\token_cache.bin. After this, the server refreshes access tokens silently on every call — no re-login needed.

To switch accounts later, just run --login again and pick the other account.

4. Test it standalone

python -c "import server; print(server.list_workspaces())"

You should get a JSON list of your workspaces (each with an id and name).

5. Register with your MCP host

Point your client at server.py (see examples/mcp.json). Use the venv's Python so dependencies resolve:

{
  "mcpServers": {
    "powerbi": {
      "command": "C:\\path\\to\\powerbi-mcp\\.venv\\Scripts\\python.exe",
      "args": ["C:\\path\\to\\powerbi-mcp\\server.py"],
      "env": { "POWERBI_TENANT_ID": "your-tenant-id-or-domain" }
    }
  }
}

Restart your MCP client, then try: "using powerbi, list my workspaces".

Use with Claude Desktop

Claude Desktop reads its MCP servers from claude_desktop_config.json. Open it from Settings → Developer → Edit Config (this creates the file if it doesn't exist), or edit it directly:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Add this server under mcpServers, using absolute paths to the venv's Python and server.py:

{
  "mcpServers": {
    "powerbi": {
      "command": "C:\\path\\to\\powerbi-mcp\\.venv\\Scripts\\python.exe",
      "args": ["C:\\path\\to\\powerbi-mcp\\server.py"],
      "env": { "POWERBI_TENANT_ID": "your-tenant-id-or-domain" }
    }
  }
}

On macOS the paths are POSIX, e.g. "command": "/Users/you/powerbi-mcp/.venv/bin/python". Save the file and fully quit and reopen Claude Desktop (use Quit from the tray/menu-bar icon — closing the window isn't enough). The server's tools then appear in the tools (🔌) menu of a new chat. Sign in first with python server.py --login so the token cache exists before Desktop launches it.

Usage flow

A typical sequence:

1. list_workspaces → pick a workspace id
2. list_datasets(workspace_id) → pick a dataset id
3. get_model_definition(workspace_id, dataset_id) → learn table & measure names
4. run_dax(workspace_id, dataset_id, "EVALUATE ...") → get results

Security

The MSAL token cache (%LOCALAPPDATA%\powerbi-mcp\token_cache.bin) contains a refresh token. Treat it like a password — it lives outside the repo and is never committed.

License

MIT (c) 2026 Rajiv Datta