claude-ecom MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

CSV in, structured review out
KPI decomposition (revenue = traffic × CVR × AOV)
Python-backed calculations for accuracy
Prioritized findings, not a wall of charts
Next-actions list tied to findings

Démo en direct

Aperçu en pratique

claude-ecom-skill.replay ▶ prêt

0/0

Installer

Choisissez votre client

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

{
  "mcpServers": {
    "claude-ecom-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/takechanman1228/claude-ecom",
        "~/.claude/skills/claude-ecom"
      ],
      "_inferred": true
    }
  }
}

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

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "claude-ecom-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/takechanman1228/claude-ecom",
        "~/.claude/skills/claude-ecom"
      ],
      "_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": {
    "claude-ecom-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/takechanman1228/claude-ecom",
        "~/.claude/skills/claude-ecom"
      ],
      "_inferred": true
    }
  }
}

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

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "claude-ecom-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/takechanman1228/claude-ecom",
        "~/.claude/skills/claude-ecom"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "claude-ecom-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/takechanman1228/claude-ecom",
        "~/.claude/skills/claude-ecom"
      ]
    }
  ]
}

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

~/.config/zed/settings.json

{
  "context_servers": {
    "claude-ecom-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/takechanman1228/claude-ecom",
          "~/.claude/skills/claude-ecom"
        ]
      }
    }
  }
}

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

claude mcp add claude-ecom-skill -- git clone https://github.com/takechanman1228/claude-ecom ~/.claude/skills/claude-ecom

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

Cas d'usage

Usages concrets : claude-ecom

Weekly sales review in 10 minutes instead of 2 hours

👤 DTC founders, e-commerce managers ⏱ ~15 min beginner

Quand l'utiliser : Monday morning review of last week's numbers.

Prérequis

Orders CSV export from Shopify/WooCommerce/etc. — Any platform's standard order export works; column mapping is flexible
Python available locally — python3 --version; skill uses a local Python env
Skill installed — git clone https://github.com/takechanman1228/claude-ecom ~/.claude/skills/claude-ecom

Déroulement

Hand over the CSV

Use claude-ecom. Here's last week's orders.csv — do a KPI decomposition and tell me what moved.✓ Copié

→ Revenue, CVR, AOV numbers with week-over-week deltas
Ask for findings

Rank the top 3 findings by impact. Be specific — named SKUs, traffic sources, time windows.✓ Copié

→ Findings with data points, not 'revenue went up'
Extract actions

For each finding, propose one concrete action I can take this week.✓ Copié

→ Action list I could paste in a Monday standup

Résultat : A weekly review I actually read.

Pièges

Garbage CSVs produce garbage reviews — Clean the export first — merge refunds, exclude test orders

Combiner avec : filesystem

Understand why a specific SKU is underperforming

👤 Product managers, merchandisers ⏱ ~20 min intermediate

Quand l'utiliser : A hero SKU suddenly slowed down and you need to know whether it's traffic, conversion, or pricing.

Déroulement

Filter to SKU

Use claude-ecom. Focus on SKU SHIRT-BLK-M for the last 60 days. Decompose revenue into traffic × CVR × AOV and compare to the prior 60.✓ Copié

→ Decomposition with clear delta per factor
Check channel split

Is the drop concentrated in one channel or across all?✓ Copié

→ Channel breakdown with concentration analysis

Résultat : A diagnosis, not a vibe.

Customer cohort retention report from order data

👤 Growth teams watching LTV ⏱ ~30 min intermediate

Quand l'utiliser : Quarterly review of cohort repeat rates.

Déroulement

Cohort the data

Use claude-ecom. Build a monthly acquisition-cohort retention table — orders/customer by month since first order.✓ Copié

→ Triangular retention table
Compare cohorts

Which cohort is best? Worst? What's different about them?✓ Copié

→ Hypotheses grounded in the data

Résultat : A retention view you could share with an investor.

Pièges

Short data windows distort recent cohorts — Flag cohorts with <3 months of history as provisional

Combinaisons

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

claude-ecom-skill + filesystem

Point at a folder of weekly CSVs for automated trend analysis

Run the weekly review for each CSV in data/weekly/ and build a running dashboard.✓ Copié

claude-ecom-skill + bigquery-server

Instead of CSVs, pull from a warehouse

Query BigQuery for last week's orders and feed into claude-ecom for the review.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler	Coût
load_csv	path, column mapping	Starting any analysis	local
kpi_decompose	date range	Weekly / monthly review	local Python
rank_findings	analysis output	After decomposition	0
cohort_table	cohort granularity	Retention analysis	local

Coût et limites

Coût d'exécution

Quota d'API: None
Tokens par appel: Moderate — summary tables only, not raw rows
Monétaire: Free (needs local Python)
Astuce: Aggregate in Python; never paste raw orders rows into prompts.

Sécurité

Permissions, secrets, portée

Stockage des identifiants : None

Sortie de données : Order summaries and findings are sent to Claude API; raw rows can stay local if you keep aggregation server-side

Order data often includes customer PII. Mask or exclude email/name/address columns before analysis.

Dépannage

Erreurs courantes et correctifs

Python deps missing

The skill uses pandas; run pip install -r ~/.claude/skills/claude-ecom/requirements.txt

Vérifier : python -c 'import pandas'

Column mapping fails

Standard Shopify/Woo exports work out of the box. For custom exports, supply a column map in the prompt.

Wild numbers in the output

Check for duplicate order rows or non-currency values in the revenue column

Alternatives

claude-ecom vs autres

Alternative	Quand l'utiliser	Compromis
Looker Studio / Metabase dashboards	You want persistent dashboards, not one-shot reviews	Setup cost; no LLM-generated narrative
Shopify's own reports	Quick built-in views	Shallow; no cross-store or cohort analysis

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills