← 返回 Skills 市场
theoddbrick

Lobstersearch

作者 theoddbrick · GitHub ↗ · v0.1.2 · MIT-0
cross-platform ✓ 安全检测通过
234
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install lobstersearch-mcp
功能描述
AI-native agentic commerce — 10 MCP tools to discover businesses, browse product catalogs with variants, and complete purchases with Stripe payments. Built f...
使用说明 (SKILL.md)

LobsterSearch — Agentic Commerce

LobsterSearch is an agentic commerce platform. AI agents can discover businesses, browse real product catalogs with variant-level detail, and complete purchases with Stripe Connect payments — all through a single MCP endpoint.

MCP Server

Endpoint: https://mcp.lobstersearch.ai/mcp Transport: StreamableHTTP Authentication: None required (public, rate-limited)

Configuration

{
  "mcpServers": {
    "lobstersearch": {
      "url": "https://mcp.lobstersearch.ai/mcp",
      "transport": "streamable-http"
    }
  }
}

Tools (10 available)

Discovery

search

Search for businesses with natural language. Uses a 4-step fallback chain (structured → tags → LLM interpret → LLM catalog scan) for maximum recall.

Parameters:

  • query (string, required) — e.g. "best ramen in Bugis" or "dental clinic near Tanjong Rhu"
  • business_type (string) — filter by type (e.g., restaurant, salon)
  • city (string) — city name filter
  • limit (number) — 1-20, default 10

details

Get full business profile by slug or ID. Returns contact info, hours, quality scores, and catalog offerings with variants.

Parameters:

  • slug (string) — business slug (e.g., "tanjong-rhu-dental")
  • id (string) — business UUID

compare

Side-by-side comparison of 2-5 businesses.

Parameters:

  • business_ids (array, required) — 2-5 business UUIDs

Catalog

browse_catalog

Browse a business's product/service catalog. Filter by category, sort by name or price.

Parameters:

  • business_id (string, required) — business UUID
  • category (string) — filter by category
  • sort_by (string) — "name", "price_low", or "price_high"

get_product_details

Deep product view with all variant dimensions, combinations, prices, SKUs, and stock levels.

Parameters:

  • offering_id (string, required) — product/offering UUID

check_availability

Real-time stock check for a product or specific variant combination.

Parameters:

  • offering_id (string, required) — product/offering UUID
  • options (object) — variant options, e.g. {"Color":"Black","Size":"Large"}

Commerce

create_order

Create a draft order with atomic stock reservation. Items specified by offering + variant + quantity.

Parameters:

  • business_id (string, required)
  • items (array, required) — [{offering_id, variant_id?, quantity}]
  • customer_email (string) — for notifications
  • customer_name (string)
  • customer_notes (string)

confirm_order

Confirm draft order and generate Stripe Checkout payment link (30-min expiry).

Parameters:

  • order_id (string, required) — from create_order

get_order_status

Full status, payment state, fulfillment progress, and event timeline.

Parameters:

  • order_id (string, required)

cancel_order

Cancel with automatic refund (if paid) and stock restoration.

Parameters:

  • order_id (string, required)
  • reason (string)

Order Lifecycle

DRAFT → CONFIRMED → PAID → ACCEPTED → PREPARING → FULFILLED → COMPLETED

Cancellation with auto-refund available at any stage.

Key Design Principles

  • Every response includes next_actions — the agent always knows what to do next
  • Error responses include retry_guidance for agent resilience
  • Real-time stock with atomic reservation prevents overselling
  • Stripe Connect direct charges for secure payment processing
  • Businesses self-manage their data — catalog, pricing, and stock are always current

Usage Tips

  • Use search for most queries — it handles natural language and intent classification automatically
  • Use details after search to get the full profile for a specific business
  • Use browse_catalogget_product_detailscheck_availability to explore what's available
  • Use create_orderconfirm_order for the purchase flow
  • Use compare when the user wants to evaluate options side-by-side
  • Always reference next_actions in responses to guide the conversation
安全使用建议
This package is internally consistent and acts as a client-side connector to a public LobsterSearch MCP endpoint. Before installing, verify you trust the remote service (https://mcp.lobstersearch.ai) because agent calls will send order details and user contact info to that external server. Avoid sending any unrelated secrets (API keys, system tokens) through the connector. If you plan to let agents run autonomously, monitor their actions and rate-limit or restrict them while testing (use test accounts and non-sensitive data). Check the service's privacy, refund, and payment handling policies (Stripe Connect flow) and optionally confirm the GitHub repository and website match the publisher/trust signals.
功能分析
Package: lobstersearch (mcp) Version: 2.0.0 Description: AI-native agentic commerce — discover businesses, browse products, and complete purchases through MCP. Built for AI agents that shop on behalf of users. The package is a configuration-only MCP (Model Context Protocol) server definition. It contains no local executable code, scripts, or binaries. Instead, it provides configuration files (JSON, YAML) and documentation to connect AI agents to a remote commerce API hosted at lobstersearch.ai. The tools defined facilitate business discovery, product catalog browsing, and order management via Stripe. There is no evidence of local execution, credential exfiltration, or malicious payloads.
能力评估
Purpose & Capability
Name, description, mcp.json, and SKILL.md consistently describe a public LobsterSearch MCP server offering 10 tools for discovery, catalog browsing, and order lifecycle management. Required binaries (curl) and included example files align with a client-side connector/usage guide — nothing unnecessary is requested.
Instruction Scope
SKILL.md only instructs clients/agents how to call the remote MCP endpoint and how to run flows (search → details → browse → create/confirm order). It does not instruct reading local secrets or unrelated files. Note: agent use will send user-supplied PII/order details (email, name, order items) and receive actionable next_actions from the remote server — this is expected for an ordering service but is a privacy and trust surface the user should consider.
Install Mechanism
Instruction-only skill with no install spec and only small example files; nothing is downloaded or written to disk by the skill package itself. Low installation risk.
Credentials
The skill requests no environment variables, credentials, or config paths. The operations described (public endpoint, Stripe Checkout links) do not require the user's secrets. This is proportionate to the stated functionality.
Persistence & Privilege
always is false, skill is user-invocable and allows normal autonomous invocation (platform default). The skill does not request permanent agent-level privileges or modify other skills/config — appropriate for a connector to an external service.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install lobstersearch-mcp
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /lobstersearch-mcp 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v0.1.2
Version 2.0.0 is a major upgrade introducing agentic commerce capabilities and an expanded set of tools. - Expanded from 7 to 10 MCP tools, adding catalog browsing and full order/payment flow. - Enables AI-native shopping: businesses, real product catalogs (with variants), and full commerce (order, Stripe payment, cancellation). - New tools include browse_catalog, get_product_details, check_availability, create_order, confirm_order, get_order_status, and cancel_order. - Unified natural language "search" with advanced fallback and richer discovery options. - Every response now includes next_actions for agent guidance and error cases provide retry_guidance. - Updated documentation reflecting a focus on agentic commerce and real-time business data.
v0.1.1
lobstersearch-mcp v0.1.1 - Updated documentation to state worldwide business category coverage (not limited to Singapore, UK, US, Australia, and Malaysia). - Clarified description for the country_code parameter in search_businesses. - No functional/API changes; improvements are limited to documentation for accuracy and clarity.
v0.1.0
Initial release of LobsterSearch MCP skill. - Provides AI-optimized search for 46 local business categories across 5 countries with detailed structured data. - Includes 7 tools: business search, business details, nearby finder, promotions, trending, comparison, and data mismatch reporting. - Offers advanced data fields: services with pricing, products, promotions, hours, payment methods, and GEO/AI scores. - MCP server supports real-time, streamable business queries with no authentication required. - Alternative REST API and OpenAPI specification also available for integration.
元数据
Slug lobstersearch-mcp
版本 0.1.2
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

Lobstersearch 是什么?

AI-native agentic commerce — 10 MCP tools to discover businesses, browse product catalogs with variants, and complete purchases with Stripe payments. Built f... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 234 次。

如何安装 Lobstersearch?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install lobstersearch-mcp」即可一键安装,无需额外配置。

Lobstersearch 是免费的吗?

是的,Lobstersearch 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Lobstersearch 支持哪些平台?

Lobstersearch 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Lobstersearch?

由 theoddbrick(@theoddbrick)开发并维护,当前版本 v0.1.2。

💬 留言讨论