Packrift MCP Server
Enables AI agents to search a Shopify packaging-supplies catalog, get live pricing and inventory, recommend boxes, estimate shipping, and generate checkout URLs.
README
Packrift MCP Server
Production MCP (Model Context Protocol) server for Packrift, a Shopify packaging-supplies store. Lets AI agents search the catalog, get live pricing and inventory, recommend a box for an item, estimate shipping, and produce checkout URLs.
- Stack: Cloudflare Workers, TypeScript (strict), Hono, Zod, Streamable HTTP transport
- Backing: Shopify Admin GraphQL API (
2025-04), storepackrift.myshopify.com - Endpoint:
POST /mcp,GET /mcp(SSE),GET /(health),GET /.well-known/mcp/server-card.json
Tools
| Tool | Purpose |
|---|---|
search_products(query, limit?) |
Keyword search across the catalog. Cached 5 min in KV. |
get_product(handle) |
Full product detail incl. variants, dimensions metafield, weight. |
get_pricing(variant_ids[], quantity?) |
Live unit price + line total. Never cached. |
check_inventory(variant_ids[]) |
Live inventory count. Never cached. |
recommend_packaging(dims, weight, use_case) |
Up to 5 ranked variant suggestions with 0.5–2 in padding. |
get_shipping_estimate(zip, country, items[]) |
Carrier rates via Shopify draftOrderCalculate. |
create_cart_url(items[], discount_code?, ref?) |
Builds packrift.com/cart/...?ref=mcp[&discount=...]. |
Local development
Install Node 24+ and the deps:
cd ~/Downloads/packrift-mcp-server
npm install
Local secrets — already created at .dev.vars (gitignored):
SHOPIFY_PACKRIFT_TOKEN=shpat_...
Run the server:
npx wrangler dev --port 8787 --local
Smoke-test the MCP endpoint with curl:
# initialize
curl -s -X POST http://127.0.0.1:8787/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"curl","version":"1"}}}'
# list tools
curl -s -X POST http://127.0.0.1:8787/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'
# search
curl -s -X POST http://127.0.0.1:8787/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"search_products","arguments":{"query":"poly mailer","limit":3}}}'
Type-check:
npx tsc --noEmit
Deployment
The Cloudflare account is being created in a separate process. Once it's ready and wrangler is logged in (wrangler login), run these in order:
cd ~/Downloads/packrift-mcp-server
# 1. Create the KV namespace and copy the printed id into wrangler.toml
# (replace both `id` and `preview_id` with the same value).
npx wrangler kv namespace create CATALOG_CACHE
# 2. Set the Shopify Admin token as a secret (paste shpat_... when prompted).
npx wrangler secret put SHOPIFY_PACKRIFT_TOKEN
# 3. Deploy. Initial deploy puts the worker on
# https://packrift-mcp.<account>.workers.dev
npx wrangler deploy
# 4. (Once mcp.packrift.com is CNAME'd to the worker) uncomment the [[routes]]
# block in wrangler.toml and redeploy.
npx wrangler deploy
After deploy, the MCP endpoint is https://packrift-mcp.<account>.workers.dev/mcp (and later https://mcp.packrift.com/mcp). The server card lives at /.well-known/mcp/server-card.json.
Design notes / caveats
cartCreateis a Storefront API mutation, not Admin. The brief asked forcartCreate+cartBuyerIdentityUpdatefor shipping rates, but those don't exist on the Admin GraphQL API this server uses. The supported Admin path isdraftOrderCalculate, which is whatget_shipping_estimateuses. It returns the same carrier-rate data without creating a real order.- Dimensions parsing. Packrift product dimensions live in
custom.specN_valuemetafields where the matchingcustom.specN_namesays "Dimensions" or "Size". Format is human-readable (12 1/8" L x 11 5/8" W x 2 5/8" H).src/dimensions.tsparses mixed fractions and falls back to scanning the title. - Recommend collections. The brief mentioned
mailer-boxescollection — that handle does not exist on the live store. We usemailers-envelopes,boxes-mailers,corrugated-boxes,bubble-wrap-foam,cushioning, andecommerce-fulfillment(verified viacollectionsquery 2026-04-29). - Use case mapping is in
src/tools/recommend_packaging.ts(COLLECTIONS_BY_USE_CASE). - Shipping rate
handlein the response is a long opaque JWT-style string — that's how Shopify returns rate handles; pass it through to subsequent calls if needed. - Errors: tool exceptions are returned as
{ content: [...], isError: true }per the MCP spec, not as JSON-RPC-3260xerrors. Protocol-level errors (unknown tool, bad JSON) do return JSON-RPC errors.
File map
src/
index.ts Hono app + MCP JSON-RPC dispatcher
shopify.ts Admin GraphQL client + id helpers
dimensions.ts Spec-string -> structured dimensions
server-card.ts /.well-known card
tools/
search_products.ts
get_product.ts
get_pricing.ts
check_inventory.ts
recommend_packaging.ts
get_shipping_estimate.ts
create_cart_url.ts
wrangler.toml Worker config (KV binding, vars, route)
package.json
tsconfig.json
.dev.vars Local-only secrets (gitignored)
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。