Merchant Web MCP
Provides BigCommerce merchants with tools for AI agents to search products, check variant stock, apply promotions, manage cart sessions with 15-minute inventory reservations, create signed checkout sessions, and access store policies, along with agent intent analytics.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Merchant Web MCPsearch for waterproof jackets under $150 and add one to my cart"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Merchant Web MCP β Agent Analytics & Operability Prototype
Turn any e-commerce storefront into an observable, measurable, and transactable platform for AI shopping agents.
This repository holds two apps on the same agentic-commerce ground. This document is the merchant cockpit. The seller-side product thesis stays in its own briefing.
App | Role | Start |
1. Merchant Web MCP cockpit | Observable storefront MCP for AI shopping agents (this README) |
|
2. Seller Agent Console | PSP control plane for managed seller agents |
|
Prototype & Simulation Notice: This repository contains an interactive simulation cockpit and prototype of the proposed BigCommerce / Merchant Web MCP plugin architecture. The simulation engine runs client-side with simulated latency, a labelled demo checkout handoff (not a real signature), real 15-minute TTL inventory reservation timers, and an executive Agent Intent Analytics dashboard.
π― The Core Problem & Value Proposition
The Near-Term Reality: Blindness to Agent Traffic
Today, merchants are blind to AI shopping agents (ChatGPT Agent, Claude, Perplexity, Apple Intelligence). When bots scrape product pages:
Traffic is indistinguishable from dumb web scrapers.
Visual scraping consumes 6,000β10,000 tokens per page and frequently breaks on React hydration or DOM redesigns.
Merchants have zero visibility into what products agents are searching for, what queries fail, and what commercial demand passes through uncaptured.
The Solution: BigCommerce Merchant Web MCP
Agent Intent Analytics (Immediate SaaS Value): Captures incoming agent sessions, tool executions, discovered intent value (Β£124k+ pipeline), provider breakdown, and catalog demand gaps.
Deterministic MCP Interface: Exposes structured JSON-RPC tools (
search_products,get_product_details,check_variant_stock,apply_promotions,add_to_cart_session,create_checkout_session).Demo checkout handoff: The cockpit labels the session
DEMO ONLY - not signed. Real HMAC-SHA256 signing runs only in the Cloudflare Worker.
Related MCP server: Shopify Agentic MCP Gateway
ποΈ Architecture Overview
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β AI Shopping Agent β
β (Claude / ChatGPT / Perplexity) β
βββββββββββββββββββββββββββββ¬βββββββββββββββββββββββββββββ
β (1) Discover: /.well-known/mcp.json
β (2) Transport: HTTP / SSE Stream
βΌ
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Merchant Web MCP Plugin Engine β
β ββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β Layered Security Guard (Unicode Filter & Regex) β β
β ββββββββββββββββββββββββββ¬ββββββββββββββββββββββββββ β
β β β
β ββββββββββββββββββββββββββΌββββββββββββββββββββββββββ β
β β Standardized MCP Tools: β β
β β β’ search_products(query, max_price, waterproof) β β
β β β’ check_variant_stock(variant_id, postal_code) β β
β β β’ apply_promotions(promo_code, subtotal) β β
β β β’ add_to_cart_session(variant_id, qty) [15m TTL]β β
β β β’ create_checkout_session (demo label in cockpit)β β
β β β’ get_store_policies() β β
β ββββββββββββββββββββββββββ¬ββββββββββββββββββββββββββ β
βββββββββββββββββββββββββββββΌβββββββββββββββββββββββββββββ
β
βββββββββββββββββ΄ββββββββββββββββ
βΌ βΌ
βββββββββββββββββββββββββ βββββββββββββββββββββββββ
β Merchant Catalog / DB β β Storefront Cart & DOM β
β (BigCommerce / D1) β β (Live State Sync) β
βββββββββββββββββββββββββ βββββββββββββββββββββββββπ Running each app
1. Merchant Web MCP cockpit
# Install dependencies
bun install
# Test suite
bun test
# Development server (Vite)
bun run devLive deployments:
Interactive Simulation Cockpit: https://merchant-web-mcp.web.app
Standalone projector demo (single source
demos/merchant-web-mcp-demo-v2.html; Vite dev, preview, and build publish it to/demos/merchant-web-mcp-demo-v2.htmland/merchant-web-mcp-demo-v2.html): https://merchant-web-mcp.web.app/demos/merchant-web-mcp-demo-v2.html
2. Seller Agent Console
PSP-hosted control plane for managed seller agents (Mandate Gate, human approve, SP-API writes). Demo merchant: North Harbour Gear. Branding: Meridian Pay.
Product thesis, architecture, and what is simulated vs real: apps/seller-agent-console/BRIEFING.md. App notes: apps/seller-agent-console/README.md.
No build step. The prototype is static HTML/CSS/JS.
# from the repo root
bun run dev:seller
# β http://localhost:8931 (prototype/index.html)
# or open the file directly
open apps/seller-agent-console/prototype/index.htmlArchitecture diagrams: apps/seller-agent-console/prototype/architecture.html.
Imported from seller-agent-console at 3ff17182f74373d08c2cab1dbd45f40dd5db5b19 (subtree; see apps/seller-agent-console/ORIGIN.md).
π‘οΈ Security & Integrity Highlights in this Repo
Checkout HMAC stays on the Worker:
server-examples/checkoutHmac.tssigns and verifies with theMERCHANT_HMAC_SECRETWorker secret. Signing throws if that secret is unset. Verification returns false and logs. The browser cockpit never receives the secret. Its handoff is labelledDEMO ONLY - not signed.Active TTL Inventory Expiration: Implemented in
src/mcp/merchantMcpEngine.ts. Ephemeral reservations expire after 900 seconds (15 mins), automatically returning stock to available inventory and notifying the client wire HUD.Layered Defense-in-Depth: Implemented in
src/mcp/securityGuard.ts. Normalizes Unicode/zero-width obfuscation and scans across customer reviews, search queries, and promo arguments.Multi-Cart Isolation: Engine maps carts explicitly by
cart_id, rejecting checkout attempts on invalid or foreign cart sessions.
πΊοΈ Roadmap & Production Specification
For the target Cloudflare Worker + BigCommerce edge server specification, see ROADMAP.md.
This server cannot be deployed
Maintenance
Related MCP Connectors
AI shopping gateway for product search, inventory, carts, and merchant-hosted checkout.
AI-agent product catalog: search, lookup & purchase routing over verified merchant data.
The independent agent-commerce protocol for AI-agent checkout on any online store.
The E-commerce Marketplace For Every AI Agent. A Governed Catalog Of 600M+ US Products, Plus Four Agent Ready Capabilities Search, Price Comparison, Product-Vs-Product, And Detailed Q&A.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to perform e-commerce operations through a standardized interface implementing the x402 and Agentic Commerce Protocol (ACP). Supports merchant sessions, product search, and payment orchestration with production-ready security features.-
- AlicenseAqualityCmaintenanceEnables AI agents to autonomously browse inventory, negotiate terms, manage carts, and execute secure payments on Shopify stores using standardized protocols. It provides a bridge for LLMs to handle the entire commerce lifecycle from discovery to order tracking through a verifiable mandate chain.52MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to discover products, build carts, and complete purchases across multiple downstream commerce services through a secure, contract-driven API.-
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to discover and interact with a merchant's commerce layer, including product listing and checkout creation, while payments settle directly to the merchant's Stripe account.MIT