Sign In

Giggle Trade MCP Agent: Connect AI Assistants to B2B Refurbished Device Data

Published: February 15, 2026

Executive Summary

Giggle Trade exposes its B2B refurbished-device catalog, stock, cart, orders, and learning insights through a Model Context Protocol (MCP) and Agent Communication Protocol (ACP)–compliant API. Whether you use Cursor, Claude Desktop, a custom AI assistant, or any HTTP client, you can query real-time catalog listings, check stock availability, manage carts and orders (with authentication), and leverage trending and recommendation data—all through a single, documented endpoint surface.

This article introduces the Giggle Trade MCP agent: what it is, why it matters for B2B automation and AI-driven workflows, what capabilities it exposes (tools, resources, learning/insights), and how to get started with discovery, REST (ACP), and native MCP.

1. What Is the Giggle Trade MCP Agent?

The Giggle Trade MCP Agent is a standards-based interface that exposes Giggle Trade’s backend services to AI assistants and automated systems. It supports:

  • MCP (Model Context Protocol) — JSON-RPC over HTTP for native MCP clients (e.g. Cursor, Claude Desktop).
  • ACP (Agent Communication Protocol) — REST API to list and invoke the same tools without an MCP SDK.

All tools map to existing backend REST APIs (catalog, product, stock, cart, orders, addresses, logistics, payment, content, learning/insights, support). The agent adds a unified description layer, authentication handling, and optional learning/insights tools so that both humans and agents can discover and call the same capabilities.

2. Why It Matters for B2B and AI Workflows

Unified Access for Humans and Agents

Procurement and IT teams often need to:

  • Compare catalog listings and real-time stock.
  • Check product details, grades, and pricing.
  • Build carts, convert to orders, and track logistics.

The same actions should be available to AI assistants and scripts. The MCP/ACP layer ensures one contract: the same tools and parameters whether invoked from a dashboard, a script, or an AI model.

Learning and Insights in the Loop

Beyond static catalog and stock data, the agent exposes learning/insights tools—trending models, related products, and personalized recommendations—that power the website’s “Recommended” sort and “Suggested for you” in the cart. Automations and AI can use the same signals for smarter suggestions and reports.

Standards-Based and Documented

Discovery is via GET /.well-known/agent.json: agents learn the agent name, version, protocols (MCP and ACP), and endpoint URLs. Tool list and parameters are described so that clients can generate UIs or prompts from the same source. OpenAPI and architecture docs align with these capabilities for consistency and regression testing.

3. Capabilities at a Glance

Tools (38 total)

  • Public (8): Catalog options, models, list; product detail and stock; stock list and detail; content offers. No authentication required.
  • Learning / Insights (3): get_trending_models, get_related_products, get_personalized_suggestions — back the catalog “Recommended” sort and cart suggestions.
  • Authenticated (30): Cart (get, add, update, delete, batch, submit, convert to order); orders (list, get, cancel, confirm, process, ship, execution); addresses; stock (availability, authenticated list); logistics; payment (deposit, balance); messages; locations. Require JWT or API Key.

Resources (4)

Read-only reference data for agents and prompts:

  • secondhand://glossary — Business terms (model, grade, lot, order, B2B, logistics).
  • secondhand://grades — Device grade definitions (A+, A, B, C, A+++).
  • secondhand://b2b_regions — Regions and compliance (e.g. HK hub, EU VAT, DDP/DDU).
  • secondhand://api — Summary of API and MCP capabilities.

Prompts (6)

Predefined prompt snippets for common tasks (e.g. stock overview, catalog vs stock comparison, B2B selection advice, regional delivery overview). Clients can use these to seed assistant behavior.

4. How to Use: Discovery, ACP REST, and MCP

Step 1: Discovery — GET /.well-known/agent.json

No authentication. Returns the agent name, version, protocols, capabilities (including learning/insights), and endpoint URLs (MCP, ACP, health).

Example (replace with your deployment origin):

GET https://giggletrade.com/.well-known/agent.json

Use this to discover base URLs and whether learning/insights are available.

Step 2: List Tools — GET /acp/tools

Returns all tools with name, description, and input schema. Cache-friendly (e.g. 5 minutes). No auth required for the list itself.

Step 3: Invoke a Tool — POST /acp/tools/:toolName

Send a JSON body with the tool’s arguments. For authenticated tools, send:

  • Authorization: Bearer <JWT> or
  • X-Api-Key: <API_KEY>

Response includes status, result, and metadata (tool name, timestamp, request ID). Timeout and body size limits apply (see developer docs).

Step 4: MCP-Native Clients

For Cursor, Claude Desktop, or other MCP clients, use the MCP endpoint URL from agent.json (e.g. https://giggletrade.com/mcp). Connect with your client’s MCP configuration; the same tools and resources are available via JSON-RPC (tools/list, tools/call, etc.).

5. Learning and Insights: What Agents Can Do

The Learning / Insights tools expose the same data that powers the website’s recommendation and suggestion features:

  • get_trending_models — Trending entities by period and dimension (model, product, category), based on view/add-to-cart/purchase signals.
  • get_related_products — Products related to a given product (e.g. co-view, co-cart), used on product detail and “Recommended for you.”
  • get_personalized_suggestions — Personalized or global recommendations; used for catalog “Recommended” sort and similar flows.

These tools are documented in the developer MCP page and in the agent’s capability description. They support optional parameters (e.g. period, limit, experiment) for A/B or strategy variants where available.

6. Security and Authentication

  • Public tools — Catalog, product, stock list/detail, content offers, and learning/insights read APIs do not require authentication.
  • Authenticated tools — Cart, orders, addresses, payment, logistics, messages, and authenticated stock require a valid JWT (obtained via login or API Key exchange) or API Key in the request.
  • Best practice — Do not ship API keys or tokens in client-side or public code. Use server-side or secure credential storage and pass tokens only in headers.

Details and token-by-API-key flow are described in the MCP & ACP API page and backend OpenAPI documentation.

7. Where to Learn More and Try It

  • Developer portal: MCP & ACP API — Tool list by category, Learning/Insights subsection, endpoints, and copy-paste examples.
  • Discovery: GET /.well-known/agent.json on your deployment origin.
  • OpenAPI: Backend docs/openapi.yaml includes learning, insights, and support paths for contract and integration testing.

Conclusion

The Giggle Trade MCP agent provides a single, standards-based way for AI assistants and automation to access catalog, stock, cart, orders, and learning insights. By supporting both MCP and ACP and documenting capabilities in agent.json and the developer portal, Giggle Trade keeps human and agent workflows aligned and easier to integrate, test, and extend.

Next Steps

  1. Discover: Call GET /.well-known/agent.json and note the MCP and ACP URLs.
  2. List tools: Call GET /acp/tools and inspect tool names and parameters.
  3. Try a public tool: Invoke e.g. get_catalog_list or get_trending_models via POST /acp/tools/<toolName> with an empty or minimal body.
  4. Add auth for full flows: Use login or API Key to obtain a JWT, then call cart, order, or payment tools as needed.
  5. Integrate your client: Point Cursor, Claude, or your app at the MCP or ACP endpoint and build workflows on top of the same capabilities the website uses.

About Giggle Trade

Giggle Trade is a B2B platform for wholesale refurbished smartphones, offering catalog, stock, cart, and order management with quality assurance and competitive pricing. The MCP/ACP layer extends these capabilities to AI assistants and developers for integration, automation, and data-driven procurement.

For integration support or enterprise API access, contact the Giggle Trade team or visit the Developer section.

Related Resources

Disclaimer: This article describes the Giggle Trade MCP/ACP interface and typical use cases. Endpoint URLs, tool names, and authentication details may be updated over time; always refer to the current /.well-known/agent.json and developer documentation for your deployment.

Share:

Explore Giggle Trade

Browse Catalog View Stock Weekly Prices Device Grades Contact Us

Related Articles

1978 Devices, Resource Devices, BS Machines, and CPO — Explained

A clear guide to understanding the different categories of Apple refurbished devices: 1978 devices, resource machines (BS machines), and Apple Certified Pre-Owned (CPO). Essential knowledge for B2B buyers sourcing used Apple products.

Knowledge Base1/15/2026

The 7-Day Procurement Rhythm: A Practical Weekly Operating Model for B2B Phone Buyers

A weekly sourcing framework for used and refurbished phone buyers to improve inventory turns, reduce pricing mistakes, and keep quality risk under control.

Operations & Procurement4/7/2026

The Anatomy of a Grade: Deconstructing Quality Standards for B2B Buyers

A practical framework for IT and procurement teams to match different device grades (A, B, C) with specific employee roles, optimizing productivity and Total Cost of Ownership (TCO) in enterprise device deployment strategies.

Business Strategy1/19/2026