AIGC: Label: "1" ContentProducer: 001191440300708461136T1XGW3 ProduceID: a60f8e46820ea466ef0868a20e79d739_6fc44ac2d27f11f196c5525400248c00 ReservedCode1: Ms0XkeFrR7dbnIUJDH7Za5ErwbcaOgxsDCu7djCtv5/ImJ0CLsCcutypQRUaR9Jje3M7wrn05kZ/3fpjC+TDxc44qLdP2G+oc9ciNFxXW5eK42wyCXNXF1ywPniW72FazYkwVBmTTv+ZWf5w3iZVQTEmJHUCWZXNeVFkEDS2Kndl6IoPHbC7SS1FU0g= ContentPropagator: 001191440300708461136T1XGW3 PropagateID: a60f8e46820ea466ef0868a20e79d739_6fc44ac2d27f11f196c5525400248c00 ReservedCode2: Ms0XkeFrR7dbnIUJDH7Za5ErwbcaOgxsDCu7djCtv5/ImJ0CLsCcutypQRUaR9Jje3M7wrn05kZ/3fpjC+TDxc44qLdP2G+oc9ciNFxXW5eK42wyCXNXF1ywPniW72FazYkwVBmTTv+ZWf5w3iZVQTEmJHUCWZXNeVFkEDS2Kndl6IoPHbC7SS1FU0g=

Shopify Growth MCP — AI Store Operations, SEO & Analytics

A production-grade Model Context Protocol (MCP) server for Shopify that lets AI assistants like Claude, Cursor, ChatGPT, VS Code and Codex operate your store through natural language — while giving merchants, agencies and SEO professionals actionable growth workflows built on real store data.

Stop juggling 20 tabs in Shopify Admin. Ask your AI assistant to find low-stock products, audit SEO metadata, summarize sales trends, or draft marketing copy — and get structured, safe, reviewable results.

• 61 MCP tools across products, orders, customers, collections, inventory, metafields, tags, SEO, GEO, analytics, feeds and marketing.

• Read-only by default, with plan previews, bound approvals and audit logging for every write.

• Runs with zero configuration in demo mode (fictional store data) — no Shopify credentials required to explore.

• Deterministic core: Shopify connectivity and analytics never require an LLM API key.

Related MCP server: shopify

Why Shopify Growth MCP?

Most Shopify MCP servers are thin API wrappers: they expose a CRUD surface but don't help you decide what to do next. Shopify Growth MCP combines reliable store operations with a growth layer that turns raw Shopify data into prioritized, explainable recommendations:

Feature Highlights

Core store operations (61 tools)

• Products — list, search, create, update, archive, delete, media, variants, metafields, tags.

• Orders — list, get, update, cancel, draft orders, customer order history.

• Customers — list, search, get, create, update, addresses, order summaries (privacy-minimized).

• Collections — list, get, create, update, membership management, uncategorized-product detection.

• Inventory — levels, locations, set quantities, low-stock alerts.

• Metafields & tags — list, set, delete, add/remove tags with namespace/key validation.

Growth differentiators

• SEO audit (seo_audit) — scans products and collections for missing titles, weak descriptions, duplicate content, missing alt text, metadata gaps; returns severity, evidence and estimated effort.

• AI-assisted SEO proposals (seo_proposals, seo_proposals_apply) — generates reviewable title/description/metadata proposals in bulk; optional LLM provider for richer drafts (template fallback included).

• GEO / AI-search readiness (geo_ai_readiness) — evidence-based audit of content clarity, naming consistency, structured data and machine readability. No ranking guarantees, ever.

• Store analytics (analytics_*) — sales totals, order counts, AOV, best sellers, sales trends, discounts, refunds, inventory exposure, product comparisons. Deterministic formulas documented.

• Product feed quality (feed_quality_audit, feed_export) — validates data against marketing-readiness criteria and exports structured feeds.

• Inventory intelligence (inventory_intelligence) — configurable low-stock thresholds, stale-product detection, reorder suggestions, variant inconsistencies.

• Marketing preparation (marketing_*) — ad copy drafts, campaign briefs, campaign naming, product grouping by category/priority.

Safety first

• Read-only by default: every tool declares a risk tier (READ / WRITE / DESTRUCTIVE).

• Plan-and-apply workflow: bulk writes produce a preview (IDs, current values, proposed values, affected counts, warnings); an approval token is bound to the exact plan hash — changed parameters require a new approval.

• Audit logging: every operation is recorded with redacted secrets.

• Least-privilege guidance: setup docs show the minimal Admin API scopes.

• No silent writes, no fabricated results, no fake "approved by Google/Meta" claims.

Architecture

┌──────────────────────────────────────────────────────────────────────┐
│  MCP clients: Claude Desktop · Claude Code · Cursor · VS Code · Codex │
└──────────────────────────────────────────┬───────────────────────────┘
                                           │ Model Context Protocol (stdio)
┌──────────────────────────────────────────▼───────────────────────────┐
│                       shopify-growth-mcp (this repo)                 │
│  ┌───────────────┐  ┌──────────────┐  ┌───────────────────────────┐  │
│  │ Tool registry │  │ Policy engine│  │ Audit log / approvals     │  │
│  │ 61 tools      │  │ READ/WRITE/  │  │ previews, bound tokens,   │  │
│  │ risk-tiered   │  │ DESTRUCTIVE  │  │ redaction                 │  │
│  └──────┬────────┘  └──────┬───────┘  └───────────────────────────┘  │
│         │                  │                                         │
│  ┌──────▼──────────────────▼──────────────────────────────────────┐  │
│  │ Shopify GraphQL client: pagination · retries · rate limits ·   │  │
│  │ timeouts · cost awareness · API version pinning (2025-10)      │  │
│  └──────────────────────────────┬────────────────────────────────┘  │
└─────────────────────────────────┼───────────────────────────────────┘
                                  │ Shopify Admin GraphQL API
                        ┌─────────▼─────────┐
                        │  Your Shopify store│
                        └───────────────────┘

Modules: src/server (lifecycle/transport) · src/tools (registry) · src/shopify (GraphQL client, operations, pagination) · src/auth (token lifecycle) · src/security (policy, approvals, previews, audit, redaction) · src/seo (audits, GEO, proposals, feed quality) · src/analytics (metrics) · src/workflows (inventory intelligence, marketing, store ops, plan) · src/demo (fictional demo data) · src/config (env validation) · src/errors (normalized errors) · src/utils.

See docs/architecture.md for the full design, data-flow and decision log.

Quick Start

Prerequisites

• Node.js 22.12+ (tested on Node 24).

• No Shopify account required for demo mode.

30-second demo (no credentials)

git clone https://github.com/hamid-digitalmarketer/shopify-growth-mcp.git
cd shopify-growth-mcp
npm install
npm run build
npx shopify-growth-mcp --demo

The server starts in demo mode against clearly-labeled fictional store data. Point any MCP client at it (see docs/client-configuration.md) and ask:

"Run a store health check, then show me the top 5 best-selling products and any products with low stock."

Demo data is synthetic and always labeled demo: true — it is never silently substituted for a real connection.

Connect a real Shopify store

1. Create a custom app in Shopify Admin → Settings → Apps and sales channels → Develop apps.

2. Grant the minimum scopes you need (e.g. read_products, write_products, read_orders, read_customers, read_inventory).

3. Install the app and copy the Admin API access token.

4. Configure:

cp .env.example .env
# Edit .env:
SHOPIFY_SHOP_DOMAIN=your-store.myshopify.com
SHOPIFY_ACCESS_TOKEN=shpat_xxxx

5. Start:

npx shopify-growth-mcp

Run npx shopify-growth-mcp --help for all options. See docs/setup.md and docs/troubleshooting.md for details.

MCP Client Configuration

Claude Desktop

{
  "mcpServers": {
    "shopify-growth-mcp": {
      "command": "npx",
      "args": ["shopify-growth-mcp", "--demo"]
    }
  }
}

Claude Code

claude mcp add shopify-growth-mcp -- npx shopify-growth-mcp --demo

Cursor

{
  "mcpServers": {
    "shopify-growth-mcp": {
      "command": "npx",
      "args": ["shopify-growth-mcp", "--demo"]
    }
  }
}

See docs/client-configuration.md for VS Code, Codex, environment overrides and live-mode examples.

Tool Reference

Full schemas, risk tiers and examples: docs/tools.md.

Security Model

• Risk tiers: every tool declares READ, WRITE or DESTRUCTIVE. Handlers consult a policy map and block unauthorized operations.

• Plan-and-apply: preview=true returns current vs proposed values and affected counts; applying requires an approval token bound to the plan hash.

• Destructive gates: destructive tools (e.g. product_delete, order_cancel) always require explicit confirmation — never silently executed.

• Audit trail: operations are recorded in .audit/ with secret redaction.

• Secret handling: credentials come from environment only; logs, errors and responses are redacted. See SECURITY.md and docs/threat-model.md.

Testing & Development

npm install
npm run typecheck    # strict TypeScript check
npm run lint         # ESLint
npm test             # 64 unit + integration tests (Vitest)
npm run build        # compile to dist/

CI runs typecheck, lint, tests and build on every push: .github/workflows/ci.yml.

Competitive Landscape (verified 2026-10-26)

Honest positioning against the ecosystem we studied:

We are not claiming superiority in tool count or features — we focus on practical growth outcomes (SEO audits, analytics, inventory intelligence, feed quality) with a safety-first write path. Verify everything yourself: the code, tests and research notes are all in this repository.

Roadmap

See research/roadmap.md. Near-term priorities:

1. Google Search Console integration for indexability evidence.

2. Google Analytics 4 + Google Ads / Meta Ads connectors (authorized APIs only).

3. Scheduled audits and report exports (client-ready deliverables for agencies).

4. Streamable HTTP transport for remote deployments.

5. Web dashboard for visual reports.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md before opening an issue or PR. This project follows a safety-first review process: new write tools must include policy tiers, tests and documentation.

License

MIT — see LICENSE.

Recommended GitHub Topics

shopify, mcp, mcp-server, model-context-protocol, shopify-api, ecommerce, seo, geo, ai, llm, graphql, claude, cursor, typescript, analytics, inventory, digital-marketing

Built with real store outcomes in mind: SEO audits, actionable analytics, inventory intelligence and safe automation — no fabricated features, no fake metrics. （内容由AI生成，仅供参考）