CommerceBridge
MCP server providing unified access to eBay, Shopify, Best Buy, and Etsy marketplace data via the Commerce API, enabling product searches and listing retrieval across multiple retailers.
README
CommerceBridge
A remote MCP server that wraps the live Commerce API — eBay, Shopify, Best Buy, and Etsy marketplace data behind one /v1 contract — as MCP tools any agent (Claude, ChatGPT, etc.) can call directly.
Live: https://commercebridge-isaiahduprees-projects.vercel.app/mcp (Vercel project commercebridge, team isaiahduprees-projects) — 14 tools, verified against real production data (Shopify live; eBay/Best Buy/Etsy honestly gated until their upstream app keys are provisioned). Naming note: the clean commercebridge.vercel.app alias is owned by an unrelated third-party product (a WhatsApp-commerce chatbot) — use the team-scoped URL above, same situation as commerce-api's alias collision.
CommerceBridge does no scraping and holds no data of its own: every tool is a thin, faithful fetch relay to the upstream Commerce API. Whatever that API returns comes back unmodified, including its errors.
Why Authentication: None
This connector's underlying data — eBay/Shopify/Best Buy/Etsy marketplace listings — is fully public and carries zero per-user state. There is no account to log into, nothing to authorize, and nothing to bill per-seat. So, deliberately, CommerceBridge has:
- No OAuth — no login flow, no token exchange, no
/.well-known/oauth-*endpoints - No Supabase — no user table, no session storage
- No billing — no plans, no usage metering, no upgrade gate
This is the same architectural call this org made for StoreBridge, its sibling connector over App Store/Google Play data: public read-only marketplace data gets the lean, stateless build. It is deliberately leaner than this org's OAuth-based Bridge-family connectors (RemindersBridge, MessagesBridge, etc.), which proxy a signed-in user's own private data and do need auth.
Tools
14 tools, one per Commerce API endpoint, grouped by marketplace/engine:
eBay — needs EBAY_CLIENT_ID / EBAY_CLIENT_SECRET on the upstream deployment
| Tool | Endpoint | What it does |
|---|---|---|
search_ebay |
GET /v1/search |
Search live eBay listings by keyword/category across 8 sites (US, GB, DE, AU, CA, FR, IT, ES) |
get_ebay_item |
GET /v1/item |
Get one eBay item's full detail by REST id or legacy id |
get_ebay_items |
GET /v1/items |
Bulk-fetch up to 20 eBay items in one call |
search_ebay_sold |
GET /v1/sold |
Sold/completed-item comps — needs eBay's separately-gated Marketplace Insights API |
Shopify — keyless, live right now
| Tool | Endpoint | What it does |
|---|---|---|
search_shopify_products |
GET /v1/shopify/products |
List any public Shopify store's catalog from its own /products.json |
get_shopify_product |
GET /v1/shopify/product |
Get one product by handle |
get_shopify_collections |
GET /v1/shopify/collections |
List a store's collections |
search_shopify |
GET /v1/shopify/search |
Keyword-filter a store's catalog (title/vendor/type/tags) |
Best Buy — not yet live upstream (route pending redeploy, then needs BESTBUY_API_KEY)
| Tool | Endpoint | What it does |
|---|---|---|
search_bestbuy |
GET /v1/bestbuy/search |
Search Best Buy's US catalog by keyword/category |
get_bestbuy_product |
GET /v1/bestbuy/product |
Get one product by SKU |
Etsy — not yet live upstream (route pending redeploy, then needs ETSY_KEYSTRING + ETSY_SHARED_SECRET)
| Tool | Endpoint | What it does |
|---|---|---|
search_etsy |
GET /v1/etsy/search |
Site-wide active-listing keyword search |
get_etsy_listing |
GET /v1/etsy/listing |
Get one listing (only endpoint where views is populated) |
get_etsy_shop |
GET /v1/etsy/shop |
Get one shop by shop_id or shop_name |
get_etsy_shop_listings |
GET /v1/etsy/shop/listings |
List a shop's active listings |
Every tool is annotated {readOnlyHint:true, destructiveHint:false, idempotentHint:true, openWorldHint:true} — read-only, non-destructive, idempotent, over an open/external data source.
Expected upstream errors — not a bug
Only Shopify is live today. Every other engine's tool call returns an MCP tool error (isError: true) whose text is the upstream's honest error envelope, verbatim — but the exact error differs by engine, because the upstream commerce-api production deployment currently lags its own source repo:
-
eBay (
search_ebay*,get_ebay*) — the route is deployed, but app keys aren't provisioned, so you get:credentials_required: eBay app credentials required — set EBAY_CLIENT_ID and EBAY_CLIENT_SECRET (free at developer.ebay.com). This API wraps eBay's free Browse API.search_ebay_soldadditionally always returns a501 not_enabledgate regardless of eBay keys — that endpoint needs eBay's separately-approved Marketplace Insights API, which the base Browse API keys don't grant. -
Best Buy (
search_bestbuy,get_bestbuy_product) and Etsy (search_etsy*,get_etsy*) — as of this writing, the code for these engines exists in thecommerce-apisource repo but the live production deployment has not been redeployed to include it yet, so these routes 404 at the HTTP layer, verified live:not_found: No route for GET /v1/bestbuy/search not_found: No route for GET /v1/etsy/shopOnce
commerce-apiis redeployed, these will flip tocredentials_required(BESTBUY_API_KEY/ETSY_KEYSTRING+ETSY_SHARED_SECRETmissing) until those keys are provisioned, then to live data — with zero changes needed here.
CommerceBridge deliberately does not catch, hide, or fake around any of this — the tool handler passes whatever the upstream says straight through as {isError: true, content: [{type:'text', text: '<code>: <message>'}]}, whether that's not_found, credentials_required, or eventually real data.
Configuration
| Env var | Default | Purpose |
|---|---|---|
COMMERCE_API_BASE_URL |
https://commerce-api-isaiahduprees-projects.vercel.app |
Upstream Commerce API base URL. Override to point at a staging/local Commerce API deployment. |
No other configuration. No secrets belong in this repo — API keys live only on the upstream commerce-api deployment.
Run locally
npm install
npm test # local smoke test — initialize, tools/list, 2x tools/call (see below)
Deploy
npx vercel --yes --prod
That's it — no environment variables are required for CommerceBridge itself (only the upstream commerce-api deployment needs its marketplace app keys). After deploy, point any MCP client at:
https://<your-deployment>.vercel.app/api/mcp
Local smoke test
npm test runs test/smoke.mjs, which connects a real MCP Client to buildServer() from lib/tools.js over an in-memory transport (the exact same server code api/mcp.js serves over HTTP) and checks:
initialize— server identifies itself ascommercebridgev1.0.0tools/list— all 14 tools present, each with an objectinputSchemaand the expected RO annotationstools/call→search_shopify_products(store: "allbirds.com") — assertsisErroris false and real products come back withmarketplace: "shopify"— proves live data flows end-to-endtools/call→search_ebay— assertsisErroris true and the error text containscredentials_required— proves the upstream honest gate surfaces as a clean MCP tool error, not a crash
test/local-server.mjs is a second, thinner check that runs the actual api/mcp.js / api/health.js handlers behind a plain http.Server, to confirm the Vercel-shaped handlers themselves (GET/DELETE → 405, /api/health → {ok, service, upstream}) work outside of the SDK's in-memory transport.
Architecture
lib/tools.js — buildServer(): McpServer + all 14 tool registrations + upstream fetch relay
api/mcp.js — POST /api/mcp — stateless StreamableHTTPServerTransport, no auth
api/health.js — GET /api/health — {ok, service, upstream}
test/ — smoke.mjs (in-memory client/server), local-server.mjs (HTTP handler check)
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.