mcp-engine-public MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

Reads Power BI tabular model structure (tables, columns, measures, relationships)
Runs DAX queries against the open model
Creates and modifies DAX measures
Manages relationships between tables
Local-only execution — no data leaves the machine

Démo en direct

Aperçu en pratique

engine-public.replay ▶ prêt

0/0

Installer

Choisissez votre client

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

{
  "mcpServers": {
    "engine-public": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/maxanatsko/mcp-engine-public"
      ],
      "_inferred": true
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "engine-public": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/maxanatsko/mcp-engine-public"
      ],
      "_inferred": true
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "engine-public": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/maxanatsko/mcp-engine-public"
      ],
      "_inferred": true
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "engine-public": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/maxanatsko/mcp-engine-public"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "engine-public",
      "command": "TODO",
      "args": [
        "See README: https://github.com/maxanatsko/mcp-engine-public"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json

{
  "context_servers": {
    "engine-public": {
      "command": {
        "path": "TODO",
        "args": [
          "See README: https://github.com/maxanatsko/mcp-engine-public"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add engine-public -- TODO 'See README: https://github.com/maxanatsko/mcp-engine-public'

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : mcp-engine-public

Write DAX without memorizing syntax

👤 Power BI analysts new to DAX ⏱ ~15 min intermediate

Quand l'utiliser : You know what you want in English; you don't want to debug CALCULATE for an hour.

Prérequis

Power BI Desktop with a model open — Launch Power BI Desktop and open your .pbix
MCP Engine installed and connected — See mcpengine.dev — installer under 5 min

Déroulement

Orient the model

List the tables in the open model and key measures.✓ Copié

→ Model map
Write the measure

Create a DAX measure 'YoY Revenue' that returns current revenue minus same period prior year, using Date[Date] for time intelligence.✓ Copié

→ DAX + creation confirmation
Validate

Run the measure over 2024 vs 2025 and spot-check 3 rows.✓ Copié

→ Numbers that look right

Résultat : A working measure with validation, not a guess.

Pièges

Time intelligence requires a properly marked date table — Verify Date table is marked as a date table first
Model calculations differ between DirectQuery and Import — Be aware of your storage mode; DirectQuery imposes DAX restrictions

Auto-document an existing Power BI model

👤 Analysts inheriting a .pbix ⏱ ~25 min beginner

Quand l'utiliser : The original author is gone and the model has 40 tables of mystery.

Déroulement

Dump structure

Dump the model: every table's role, every measure's DAX, every relationship.✓ Copié

→ Full structural listing
Write prose docs

Write a one-page explanation of this model for a new analyst: what it reports on, main tables, key measures.✓ Copié

→ Human-readable model doc

Résultat : Useful documentation without reading every measure yourself.

Combiner avec : filesystem

Refactor measures to follow naming + modeling conventions

👤 Teams standardizing Power BI practices ⏱ ~30 min intermediate

Quand l'utiliser : Measures are named inconsistently (e.g. 'Total Sales' vs 'SumSales').

Déroulement

Find violations

List measures that don't follow the naming convention '<Verb> <Subject> [<Modifier>]'.✓ Copié

→ Violation list
Rename safely

Propose renames; list any visuals or calculated tables that reference each old name.✓ Copié

→ Rename plan with dependencies

Résultat : A consistent measure catalog with broken references caught before you hit Apply.

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

engine-public + filesystem

Export model documentation as a Markdown file alongside the .pbix

Document the model and save as ~/Docs/PBI/sales-model.md next to sales.pbix.✓ Copié

engine-public + postgres

Cross-check PBI measures against raw Postgres queries

PBI says Q1 revenue is $4.2M. Run the same query on the source Postgres DB and tell me if there's a discrepancy.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler	Coût
get_model_structure	(none)	Session start	free (local)
run_dax	query: str	Ad-hoc DAX query	free (local)
create_measure	table, name, expression	Add new measure	free (write)
update_measure	table, name, expression	Modify existing	free (write)
manage_relationship	from_table, from_col, to_table, to_col, action: create\|delete\|update	Schema changes	free (write)

Coût et limites

Coût d'exécution

Quota d'API: None — local only
Tokens par appel: Model dumps can be large; focus queries
Monétaire: Proprietary license with permitted personal + commercial use (see mcpengine.dev/license)
Astuce: Ask for specific tables rather than a full model dump on complex reports.

Sécurité

Permissions, secrets, portée

Stockage des identifiants : None required — uses PBI Desktop's session

Sortie de données : Zero telemetry claimed; all local to PBI Desktop

Proprietary license — review terms at mcpengine.dev/license before enterprise use.
Changes write directly to the open model — save your .pbix before risky refactors.

Dépannage

Erreurs courantes et correctifs

No Power BI instance found

Ensure Power BI Desktop is running with a report open; the MCP connects to the local port it exposes.

Vérifier : Check Task Manager for msmdsrv process

DAX error: Column not found

Table/column reference is case-sensitive and must match exactly. Use the model structure dump to get names right.

Vérifier : get_model_structure

Changes don't persist after closing PBI

PBI Desktop doesn't auto-save — Ctrl+S after each batch of changes.

Vérifier : Title bar shows no asterisk

Alternatives

mcp-engine-public vs autres

Alternative	Quand l'utiliser	Compromis
Tabular Editor 3 (scripting)	You want a full scripting environment for PBI/SSAS models	Paid; no MCP surface
DAX Studio	You only need a DAX query tool	No write operations; no AI layer

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills