rangeview-mcp
Read-only MCP server for searching and browsing the Rangeview Sports product catalog, including firearms, ammunition, optics, and accessories.
README
rangeview-mcp
MCP server for read-only access to the Rangeview Sports product catalog — ~5,600 products across firearms, ammunition, optics, reloading and accessories.
Backed by the store's public WooCommerce Store API. No credentials required. Read-only by design: nothing in this server can add to a cart, place an order, or write anything.
Install
git clone https://github.com/IamMichael23/rangeview-mcp.git
cd rangeview-mcp
npm install && npm run build
Use it with Codex
Add to ~/.codex/config.toml:
[mcp_servers.rangeview]
command = "node"
args = ["/absolute/path/to/rangeview-mcp/dist/index.js"]
Use it with Claude Code
claude mcp add rangeview -- node /absolute/path/to/rangeview-mcp/dist/index.js
Or commit a .mcp.json in your project to share it with collaborators:
{
"mcpServers": {
"rangeview": {
"command": "node",
"args": ["/absolute/path/to/rangeview-mcp/dist/index.js"]
}
}
}
Tools
| Tool | Purpose |
|---|---|
search_products |
Free-text + category, caliber, brand, price, stock and sale filters |
get_product |
Full detail for one product by id, slug, or SKU |
list_categories |
Walk the 376-category tree |
list_filter_values |
Discover the 83 filterable attributes and their valid values |
check_availability |
Bulk price/stock lookup for up to 20 items |
"What bolt-action rifles in 6.5 Creedmoor do they have in stock under $2000?"
"Is SKU 20260728001 still available?"
"What calibers do they carry the most of?"
Two problems this server exists to solve
Anything that just proxies the Store API straight through will get both of these wrong.
1. A raw product is ~1,300 tokens, and 81% of it is dead weight
Measured against a live product:
| Field | Share of payload | Needed in a search result? |
|---|---|---|
description |
44.1% | No — detail view only |
images (5 sizes each) |
15.8% | No — one URL at most |
_links |
7.4% | Never |
add_to_cart |
6.3% | Never — this server is read-only |
price_html |
2.6% | No — raw cents is enough |
extensions, tags |
5.1% | Never |
So results come back in two shapes: a compact projection for search (~340 bytes / ~85 tokens) and a full shape only from get_product, with the description HTML-stripped and truncated. That is a ~15x reduction, and it is the difference between 50 results fitting in context and blowing it out. Every response also carries total_matches, page and total_pages so a model paginates deliberately instead of asking for everything.
2. Attribute filters take term IDs, and this catalog's terms are dirty
The Store API filters on term_id, never on names. And one cartridge is spelled many ways:
.223 | .223 Rem | 223 Rem | 223 Remington | .223 Rem/5.56 NATO | 5.56×45mm NATO
Inch marks arrive as two different entities for the same value — 6.5 Creedmoor/20” and 6.5 Creedmoor/20″.
Picking the single best-matching term returns 126 products for .223. ORing the whole family returns 138. So resolution deliberately fans out — you pass caliber: "223" and the server expands it to every term in the family and joins the IDs, which the Store API treats as OR. The response reports the expansion in notes.
The matching rule that makes this safe is a domain distinction: in cartridge naming a slash means "chambers both" (.223 Rem/5.56 NATO) while a hyphen means "wildcat built on that parent case" — a different cartridge that will not chamber. So 6mm-223, .17-223 and 7mm-223 Ingram are correctly excluded from a .223 search even though all three contain the string "223".
Catalog hygiene
Two quirks of the live store are handled by default, because both produce confidently wrong answers otherwise:
- 15 internal
Shipping test product #Nrows are live at $0.01. Without filtering, "what's the cheapest item?" returns a QA artifact. They are hidden by default; passinclude_test_products: trueto see them. (The cheapest genuine item is a $0.49 pack of earplugs.) $0.00means "call for price", not free. Special orders such as a Beretta 693 shotgun list at zero. These are reported asprice_on_request: truewith anullprice, so a model never announces a $2,775 shotgun is free.
Also worth knowing: only ~34% of the catalog (1,923 of 5,602) is in stock. Pass in_stock_only: true for buyable items.
Configuration
All optional:
| Variable | Default | Purpose |
|---|---|---|
RANGEVIEW_BASE_URL |
https://www.rangeviewsports.ca |
Point at another WooCommerce store |
RANGEVIEW_TIMEOUT_MS |
20000 |
Per-request timeout |
RANGEVIEW_MAX_RETRIES |
3 |
Retries on 429/5xx with backoff |
RANGEVIEW_TAXONOMY_TTL_MS |
3600000 |
Category/attribute cache lifetime |
RANGEVIEW_DESC_MAX_CHARS |
1500 |
Description truncation in get_product |
Tests
npm test # all 35
npm run test:unit # 22, offline
npm run test:integration # 13, hits the live API
Set RANGEVIEW_SKIP_INTEGRATION=1 to skip the network tests.
Notes
This reads a public, unauthenticated endpoint that the store already serves to every browser, and it is rate-limit-friendly (single-page queries, cached taxonomy, backoff on 429). It is not affiliated with or endorsed by Rangeview Sports. Product data, availability and pricing belong to them and change without notice — confirm anything that matters on the site itself.
MIT licensed.
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。