The shopping engine for AI agents, resellers, and deal-hunters

Never miss an underpriced GPU again

Shopper hunts every marketplace, forum, and local seller, ranks the deals cheapest first, and watches 24/7 with Radar. Try a real hunt right now, no sign up.

MCP tools
52
MCP tools
Radar scans
24/7
Radar scans
agents self-pay
USDC
agents self-pay
to connect
1 line
to connect

Run your own agent? Point any MCP client at Shopper in one line.

claude mcp add --transport http shopper https://shopper.sh/api/mcp

One MCP endpoint. Every agent you run becomes a shopper.

Claude Code
Cursor
Codex
OpenClaw
Hermes
any MCP client
52 tools
Radar 24/7
Deep browser
USDC-native
Claude Code
Cursor
Codex
OpenClaw
Hermes
any MCP client
52 tools
Radar 24/7
Deep browser
USDC-native

The problem

Your agent can shop. It just can't remember.

Ask an agent to find something and it dumps links into a chat, then forgets everything: no lists, no price watching, no memory between sessions. Shopper is the missing infrastructure: state, a real shopping engine, and standing watches, behind one MCP endpoint.

An agent in a chat window

  • Your agent finds links, then forgets them the moment the session ends
  • No lists: every find lives and dies inside one conversation
  • Nothing watches prices after the chat window closes
  • Every new session starts from zero, in every client

An agent on Shopper

  • Structured state: wish list, shopping lists, About You, shared by every agent
  • Every find saved with its price, seller, and source
  • Radar keeps scanning 24/7 after your agent signs off
  • One memory across Claude Code, Cursor, Codex, and any MCP client

Radar

It hunts while you sleep

Radar is a standing scan that watches the web 24/7 for exactly what you want: "recently listed pre-owned GPUs at a good price." Every match lands in your wish list, priced and sourced, ready for your agent. On paid plans.

Standing scans

Set it once and Radar keeps checking, around the clock. Fresh listings are scored against your ask, so a match means a real match, not a keyword hit.

RTX 4090, pre-owned$1,140, dropping
Walnut console table$640, steady
Espresso grinder$289, above target

Sellers vetted

Every match comes with its seller checked against public business registries, so a great price from a ghost storefront never wastes your time.

Meridian Textiles Ltdsupplier
GLEIF verifiedregistry match

The grail

Gucci loafers, 10M, under $400: Radar flags every fresh listing the moment it appears.

The project car

A clean project car within driving distance: new listings surface before the crowd finds them.

New suppliers

On Pro, point Radar at your supply chain: new manufacturers and suppliers in your niche, watched.

Price context

Every match arrives with where the price has been, so you and your agent know whether today's listing is actually a deal.

Lands in your wish list

Matches are saved as structured wish list items: price, seller, source. Any connected agent picks them up from there.

Ask about any match

You or your agent can interrogate a match: is the seller solid, is the price fair, is it the right version. Clear answers with sources, not a wall of links.

? Is this $1,140 pre-owned 4090 actually a good buy?

Yes. The seller has a long clean record, and the price sits well below where this card has traded for the last three months.

seller historyprice historylisting scan
A shopper reviewing finds on a laptop

Radar returns real listings from real sellers, never ad farms, never the wrong version. Getting it right beats casting a wide net, always.

What your agent gets

52 tools. One shopping engine.

Hunts, deep browser sessions, Radar scans, lists, memory, seller vetting, even billing: everything an agent needs to shop, over one MCP connection. Your data stays yours: exportable, never resold.

One find_items call hunts the whole web

Your agent calls find_items and Shopper fans the hunt out across stores, marketplaces, and local sellers, with search engines under the hood (Exa, Firecrawl, Tavily). Back come structured listings with price and seller, not a page of links.

find_itemspre-owned RTX 4090 under $1,400

23 listings found across 9 sites, best 4 shown

RTX 4090, pre-owned
$1,140
RTX 4090 FE, open box
$1,289
RTX 4090, refurb + warranty
$1,310
RTX 4090, local pickup
$1,050

How it works

From one line to an agent that shops

Connect once, then delegate. Your agent does the work; you stay in control of every purchase.

01

Connect your agent

One line in Claude Code, Cursor, Codex, OpenClaw, Hermes, or any MCP client: https://shopper.sh/api/mcp, with OAuth or an API key. Your agent picks up all 52 tools instantly.

https://shopper.sh/api/mcpconnected
02

Tell it what you want

One item, a whole shopping list, or a standing wish. Your agent reads About You first, so sizes, budgets, and no-gos shape the hunt from the very first call.

you find me a pre-owned RTX 4090 under $1,400
reads About Youknows your budgetchecks your lists
03

It hunts and saves the finds

The agent works stores, marketplaces, and local sellers, opens a real browser for the tricky ones, and saves every find to your wish list with its price and seller.

FindRTX 4090, pre-owned, $1,140saved
Sellerrated 4.9, registry checkedsaved
04

Radar keeps watching

Standing scans run 24/7 after the session ends, and new matches land in the wish list. If you allow it, the agent even tops up its own credits with USDC over x402.

Radar found 2 new matches. overnight

Pre-owned GPUs at a good price, freshly listed, saved to the wish list

How it fits together

One place for everything you want

You ask, Shopper searches everywhere and watches for deals, and every find lands in one wish list. That list powers your shopping lists and your price alerts, so nothing you wanted ever slips through.

It compounds

Every hunt makes the next one sharper

Every hunt grows About You, and the wish list gets smarter with each find. Because all your agents share one memory, a fact learned in Claude Code sharpens the next hunt in Cursor. This is what a month looks like when nothing gets forgotten.

Wish list items, 30 days

Illustrative view

Finds saved
312
Sellers vetted
148
Radar matches
64
Facts remembered
87

Day 30: 563 items, shared by every agent you connect, yours to export.

Why Shopper

Built foragents.

The shopping MCPs bolted onto LLM providers are stateless link dumps: no lists, no watching, no memory. Shopper is structure, an engine, standing scans, and a wallet in one system: an order of magnitude more. And the data is yours: exportable, never resold.

01

Structure, not link dumps

Bolted-on shopping MCPs return links and forget them. Shopper returns typed finds with price, seller, and source, saved into lists your agent can query, update, and check off.

02

A real engine underneath

52 tools backed by a real shopping engine: web-wide hunts, local stores, deep browser sessions for marketplaces and forums, and seller vetting against public registries.

03

It keeps watching

Stateless MCPs stop when the session ends. Radar runs standing scans 24/7 and drops matches into your wish list, so your agent wakes up to finds instead of starting over.

04

A wallet of its own

With x402, your agent buys its own credits and plans in USDC over HTTP 402, no human in the loop. No provider's bolted-on shopping tool lets an agent fund its own work.

About Shopper

Shopping is work. Agents should do it.

Shopper is the shopping engine your agents run. Connected over MCP, they hunt the whole web for items, compare prices, read reviews, vet sellers, and keep every find in lists you own.

Deal-hunting, resale sourcing, the grocery run, auto parts, a whole-home refit, or finding a manufacturer. You already run agents; Shopper gives them the engine, the state, and the memory to shop with understanding instead of search and forget.

Start free
Shopping bags and a phone

Owned data - yours to keep, yours to export

Every find, list, and About You fact lives in your Shopper account. No shared pool, no lock-in, never resold. Export it all whenever you like.

Vetted before you spend

Before a big purchase or a new supplier, sellers are checked against public registries: GLEIF, Companies House, SEC EDGAR. Trust is earned, then verified.

One memory, every client

Agents read About You before every hunt: sizes, budgets, no-gos. State persists across sessions and clients, so every agent you connect shares it.

Give your agents buying power

Connect any MCP client to https://shopper.sh/api/mcp and your agent starts hunting, saving, and watching, on lists that stay yours.

Free to start. Pro $20/mo adds sourcing; Max $49/mo runs Radar at reseller scale.