Periskop MCP Server

Periskop MCP Server

Enables AI agents to perform product discovery from natural language shopping intents, returning ranked products with merchant links without completing checkout.

Category
Visit Server

README

Periskop — Product Discovery for AI Agents

Product discovery for AI agents. One MCP call turns a natural-language shopping intent into structured, ranked product results — best picks, alternatives, bundles, caveats, prices, and merchant links — from across independent stores on the open merchant web.

Periskop is a hosted, remote MCP server (closed source). You connect over the public HTTPS endpoint with a Periskop API key — nothing to install or run. It's the discovery layer for agentic commerce: an agent asks in natural language and Periskop returns ranked products with links back to the merchant. It stops at discovery — no checkout, no payments, no orders.

  • Website: https://periskop.ai
  • Developer portal / get an API key: https://periskop.ai/developer
  • Docs: https://periskop.ai/developer/docs
  • MCP endpoint: https://mcp.periskop.ai/v1/mcp (Streamable HTTP, JSON-RPC 2.0)

Who it's for

Any product that needs to turn intent into reliable product results without owning checkout — agent platforms and orchestrators, AI worker platforms, AI shopping assistants, browser agents, commerce copilots, recommendation systems, procurement and B2B sourcing workflows, price and deal monitors, restock/reorder bots, gifting apps, resale and arbitrage workflows, and accessibility shopping tools. Concretely: an agent that plans a camping trip and returns the gear, a procurement copilot comparing options across suppliers, or a voice assistant surfacing the right product from a single sentence.

What it does (and does not) do

Periskop returns product results and merchant links only. It does not complete checkout, create merchant carts, process payment, reserve inventory, or purchase items. The user always completes any purchase on the merchant's own website. Every response carries a purchase_boundary block restating this.

Connection

Endpoint https://mcp.periskop.ai/v1/mcp
Transport Streamable HTTP — JSON-RPC 2.0 over HTTP POST (one JSON response per POST; no SSE)
Wire protocol 2024-11-05
Auth Authorization: Bearer <YOUR_PERISKOP_API_KEY> (keys look like dp_…)
Get a key https://periskop.ai/developer

Unauthenticated requests receive 401 with a WWW-Authenticate: Bearer header.

Note: Periskop also supports OAuth 2.1 (Authorization Code + PKCE) for hosts that prefer it (tokens look like dpo_…). API key is the simplest path and is all you need for the configs below.

Tools

All tools take a single JSON object argument. Schemas below are the public tool surface.

run_shopping_discovery

Find, choose, browse, recommend, get the best product, or build a bundle from a natural-language request. The only required field is prompt.

Field Type Required Notes
prompt string Natural-language shopping intent
mode string | null auto | browse | recommend | best | bundle (default auto)
store string | null auto, a store id, or a store name/hint
country string | null e.g. PT, ES
currency string | null e.g. EUR
language string | null e.g. en, pt
max_results integer | null Cap on returned products
response_format string | null full (default) | simple

get_discovery_result

Retrieve a previous result by result_id. Results are temporary and may expire.

Field Type Required
result_id string
response_format string | null

discover_supported_stores

Inspect the public stores Periskop can use. Returns public store capabilities only.

Field Type Required
country string | null
category string | null
capability string | null (search | bundle | product_links)

report_result_feedback

Report whether a result was good, bad, or mixed.

Field Type Required
result_id string
rating string (good | bad | mixed)
reason string | null
selected_product_id string | null
freeform_feedback string | null

suggest_store_coverage

Suggest a store/merchant/marketplace Periskop should support in the future. Non-billable. Does not search or scrape the store and does not guarantee future support.

Field Type Required
store_name string
store_url string | null
country / region / category string | null
context string | null (e.g. unsupported_store, no_match)

Example

Request (JSON-RPC tools/call)

curl -sS -X POST https://mcp.periskop.ai/v1/mcp \
  -H "Authorization: Bearer YOUR_PERISKOP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "run_shopping_discovery",
      "arguments": { "prompt": "wireless noise-cancelling headphones for travel under 200€", "mode": "best" }
    }
  }'

Response (trimmed — sample/fake data, not real)

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [
      {
        "type": "text",
        "text": "{\"result_id\":\"res_example_8f2a1c\",\"mode_used\":\"best\",\"products\":[{\"title\":\"SampleAudio Aero NC\",\"price\":\"€179.00\",\"currency\":\"EUR\",\"merchant\":\"Example Store\",\"url\":\"https://store.example/p/aero-nc\",\"role\":\"best_pick\"},{\"title\":\"Acme Quietline 2\",\"price\":\"€149.00\",\"currency\":\"EUR\",\"merchant\":\"Example Store\",\"url\":\"https://store.example/p/quietline-2\",\"role\":\"alternative\"}],\"caveats\":[\"Prices and availability may change on the merchant site.\"],\"purchase_boundary\":{\"checkout_created\":false,\"payment_processed\":false,\"stock_reserved\":false,\"user_must_complete_purchase_on_merchant_site\":true}}"
      }
    ]
  }
}

The structured payload is returned as text in result.content[0].text. Every response and the X-Request-ID response header carry a per-call request id.

Pricing (summary)

Prepaid wallet, billed only on successful discovery requests. Auth errors, malformed requests, internal/runtime failures, rate limits, and hard no-match are never billed. See the live rate at https://periskop.ai/developer/billing.

Client setup

Copy-paste configs are in config-snippets/: Cursor (.cursor/mcp.json), Claude Desktop, Claude Code, and a raw curl example.

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

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.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured