BuyWhere Product Search MCP
Give AI agents cross-market product search, price comparison, and deal discovery.
- Source repo
- BuyWhere/buywhere-mcp
- Stars
- ★ 14
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 55/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A shopping-assistant developer needs to find wireless earbuds within a Singapore customer's budget and return purchasable candidates.
- Not a fit if
- Teams that require fully offline product search
- Merchants needing store or Shopify management
- Environments that cannot send queries to the BuyWhere API
- Website
- buywhere.ai ↗ set on GitHub
- Maintenance
- Commits in the last 6 monthslast commit 2026-09-28latest release v0.4.0 (2026-08-30)
- Source review
- 55/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
BuyWhere MCP Server exposes BuyWhere's product catalog and real-time pricing capabilities to Model Context Protocol clients. It can run as a Node.js stdio process for Claude Desktop, Cursor, VS Code, Cline, Windsurf, OpenCode, Codex, and Continue.dev, while a hosted JSON-RPC 2.0 endpoint is available at https://api.buywhere.ai/mcp. Agents can search a catalog of more than 300 million products, retrieve details, compare 2–10 items, find the cheapest deliverable listing, discover price drops, locate similar products, and submit URLs for ingestion. Its documented coverage spans Singapore, the US, Malaysia, Thailand, and Vietnam, including Lazada, Shopee, Amazon, and local retailers. It suits shopping-research and cross-border comparison agents, but it is not a store-management system or Shopify integration, and live queries depend on BuyWhere's hosted API, network access, and an API key.
An MCP client starts @buywhere/mcp-server over stdio or connects to the remote MCP endpoint over HTTP with a Bearer credential. The server accepts structured tool calls and sends them to the product catalog at api.buywhere.ai. search_products filters by keyword, category, price, and country; get_product returns prices, specifications, and images; compare_products compares 2–10 products side by side; find_best_price finds the cheapest deliverable merchant listing; get_deals returns products with significant price drops; list_categories exposes the catalog taxonomy; find_similar performs vector-similarity discovery; and ingest_products submits product URLs for catalog ingestion. Corresponding v2 search, detail, comparison, deal, and best-price tools add delivery-country handling through deliver_to. The result is structured product, price, merchant, and delivery data that the calling agent can rank, summarize, or recommend.
- A shopping-assistant developer needs to find wireless earbuds within a Singapore customer's budget and return purchasable candidates.
- A cross-border shopper wants to compare the same product across US, Singapore, and Southeast Asian merchants.
- A research agent needs to search a category and then run
compare_productson 2–10 shortlisted items. - An automation team wants to call
get_dealsorfind_best_priceperiodically and pass price changes into another workflow. - A merchant or catalog contributor wants to submit product URLs through
ingest_productsfor BuyWhere ingestion. - A multi-agent system needs to delegate product-search and comparison tasks through the documented A2A Agent Card.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
npx -y @buywhere/mcp-servernpm install -g @buywhere/mcp-server
- Environment variables
BUYWHERE_API_KEYBUYWHERE_API_URL- Runtime
- Node.js 18
How do you install or deploy this agent?
Node.js 18 or newer is required. First obtain a free key without email verification:
curl -X POST https://api.buywhere.ai/v1/auth/register \
-H "Content-Type: application/json" \
-d '{"agent_name":"your-agent"}'The response immediately includes an api_key, tier, and rate limits. Export the key and run the MCP server directly:
export BUYWHERE_API_KEY=bw_...
npx -y @buywhere/mcp-serverA global installation is also documented:
npm install -g @buywhere/mcp-server
buywhere-mcpFor source development, use:
git clone https://github.com/BuyWhere/buywhere-mcp.git
cd buywhere-mcp
npm install
npm run build
npm startHow do you use this agent?
Add this configuration to Codex, OpenCode, or another client using the documented MCP server format:
{
"mcpServers": {
"buywhere": {
"command": "npx",
"args": ["-y", "@buywhere/mcp-server"],
"env": { "BUYWHERE_API_KEY": "bw_live_xxxx" }
}
}
}After reloading the client, ask for a concrete shopping task such as “Find wireless earbuds under $50 available in Singapore.” The client can call search_products and then pass selected IDs to compare_products. For remote operation, send JSON-RPC 2.0 requests to the hosted endpoint with the API key as a Bearer credential:
POST https://api.buywhere.ai/mcp
Authorization: Bearer bw_...The endpoint declares MCP protocolVersion 2024-11-05. BUYWHERE_API_URL can override the default API address.
What are this agent's strengths and limitations?
- One MCP interface covers Singapore, the US, Malaysia, Thailand, and Vietnam across Lazada, Shopee, Amazon, and local retailers.
- The tool set spans the full research path: discovery, product details, multi-item comparison, best price, deals, similarity, and URL ingestion.
- Agents can self-register a free key with one POST and no email; the example clearly states a 20 RPM and 1,000-per-day unverified limit.
- Both a local stdio package and a hosted MCP endpoint are documented, along with configurations for several clients and agent frameworks.
- The repository uses the MIT license and documents both MCP and A2A protocol support.
- Product operations depend on BuyWhere's hosted API; there is no evidence that the catalog or search backend can be self-hosted or used offline.
- Deployments must manage
BUYWHERE_API_KEY, and the documented unverified tier is limited to 20 requests per minute and 1,000 per day. - Explicit market coverage is limited to Singapore, the US, Malaysia, Thailand, and Vietnam.
- The documentation conflicts on tool count: the remote MCP section says 13 tools, while the protocol table says the stdio transport has 10.
- Real-time price and availability depend on upstream merchant data and BuyWhere; no refresh interval, uptime guarantee, or failure-handling commitment is stated.
How does this agent compare with similar options?
tavily-mcp has a different role: the source suggests Tavily for reviews and general product research, then BuyWhere for current prices and purchase links. @modelcontextprotocol/server-filesystem, @supabase/mcp-server-supabase, n8n-mcp, and @playwright/mcp are presented as complementary components for saving results, persisting preferences and alerts, automating monitoring, and testing browser interactions rather than direct replacements for BuyWhere's catalog search.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| BuyWhere Product Search MCP This agent | 55 · Major gaps | MCP serverFree | ★ 14 | 1mo ago | TypeScript | Codex |
| tRPC-Agent-Go | 56 · Major gaps | Library / SDKFree + model costs | ★ 1.9k | 1d ago | Go | OpenAI API |
| Fetch.ai Agent Examples | 44 · Major gaps | CLIFree + model costs | ★ 1.5k | 1d ago | Python | OpenAI API · Claude API |
| MCP Memory Service | 68 · Some gaps | MCP serverFree | ★ 2.1k | 1d ago | Python | ChatGPT · Codex · Claude Code · Claude.ai |
How does FollowAgents rate this agent?
Why each dimension lost points
The source shows a thin MCP bridge: it reads the API key from an environment variable and sends it as bearer and x-api-key headers to a configurable upstream. The README and security policy give basic key-handling guidance, supporting partial credit for least privilege, data-flow transparency, and sensitive-data handling. Deductions: calls do not require user confirmation, a custom upstream can receive the key, and ingest_products submits URLs; no revocation or recovery path is documented, dependency security has no audit evidence, and publisher identity is unverified. BuyWhere attribution is present, while maintenance contact is only partly clear.
The source handles timeouts, network errors, HTTP status, and response parsing, and CI builds on Node 18, 20, and 22. Deductions: README, Smithery metadata, and source disagree on tool counts, version contracts, registration options, and v2 status; the product depends on a hosted API without availability or fallback guarantees. Error handling helps ordinary diagnosis but does not cover every unsuccessful response or parsing case.
The README identifies MCP clients, shopping agents, price comparison, and cross-border shopping as audiences and scenarios, with multiple framework configurations and examples. Deductions: the capability is mainly product search and comparison, while data and tool contracts are controlled by the hosted service; market coverage and delivery-country requirements are inconsistent across the evidence, making trigger and environment guidance less stable.
The README provides quick installation, client configuration, tutorials, a tool table, examples, and support channels; package metadata and LICENSE specify MIT. Deductions: README, smithery.yaml, package.json, and SECURITY.md disagree on versions, tool surfaces, or supported versions; there is no clear changelog or known-limitations list in the packet. An email and issue path exist, but maintenance responsibility and support commitments are limited.
Tools cover search, details, comparisons, best price, deals, and categories, which can produce useful structured shopping information; forwarding to a hosted API avoids duplicating server logic locally. Deductions: result quality, price freshness, and market coverage depend on the external service, and the repository does not substantiate value or cost. Registration and usage-limit details also vary across files.
Core forwarding behavior, timeout handling, request headers, and tool caching are traceable in source; installation and configuration claims have partial support from package metadata, CI, and the security policy. Deductions: the supplied sample has no independent evidence for catalog size, freshness, market coverage, service quotas, or total tool count; documentation conflicts on these claims and on v2 release status, and does not consistently separate verified facts from promotional claims.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Check the hosted service's actual tools and API contract before use; repository materials conflict on tool count, v2 availability, market coverage, and quotas.
- The API key is sent to the configured upstream. Confirm BUYWHERE_MCP_URL or BUYWHERE_API_URL points to the intended service, and avoid exposing the key in client configuration or version control.
- Catalog size, live pricing, availability, and registration claims are not independently verified by this static evidence.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.4.0 v0.4.0 — stdio proxy to the hosted MCP server
Release notes excerpt
The package no longer reimplements tools. tools/list and tools/call forward to https://api.buywhere.ai/mcp and return responses verbatim, so the stdio package and the hosted server expose the same contract by construction (13 tools, including the v2 family). Previously the package hardcoded 7 v1 tools and drifted from the hosted server — an agent comparing the two saw different products.View on GitHub ↗ -
v0.3.17 v0.3.17 — Scorecard tools + description fix
Release notes excerpt
Scorecard improvement + MCP Registry description fix (<100 chars). NPM publish required.View on GitHub ↗ -
v0.3.16 v0.3.16 — Point MCP Registry to NPM 0.3.11
Release notes excerpt
Fix: align server.json with actual published NPM version 0.3.11View on GitHub ↗
FAQ
Does it require payment or an email signup?
POST /v1/auth/register with no email. Its example unverified tier allows 20 requests per minute and 1,000 per day; no paid-plan pricing is provided.Do product searches run entirely on my machine?
api.buywhere.ai, so network access is required and queries are sent to BuyWhere.Can it manage a store or Shopify catalog?
Which markets and delivery filters are supported?
deliver_to country, while v2 comparison accepts it optionally.Can I use it without installing the npm package locally?
POST https://api.buywhere.ai/mcp, using JSON-RPC 2.0 and a Bearer API key.Is BuyWhere Product Search MCP free?
Is BuyWhere Product Search MCP open source?
Is BuyWhere Product Search MCP safe to use?
Is BuyWhere Product Search MCP still maintained?
Who makes BuyWhere Product Search MCP?
What is BuyWhere Product Search MCP's official website?
Thanks — we use this to improve the page.