Skip to content

Crawl MCP

One endpoint. Any Amazon URL. Agent-ready output.

Your agent asks a product question — "what is it, what does it cost, is it in stock, what do buyers say?" — and gets the answer in a structure it can use. One call. Median 280ms. No scraper to babysit.

POST /v1/amazon/lookup
→ asin: B0CHX3QBCH
→ mode: spot_check
{
title: "Anker 737 PowerCore",
brand: "Anker",
price: "$99.99",
rating: 4.7,
reviews: 12,431,
availability: "in_stock",
asin: "B0CHX3QBCH"
}
spot_check · 1 round-trip · ~280ms median

Built for agents

Built for how agents actually work — one question, one answer.

"is B0CHX still in stock under $100?"
price$99.99
availabilityin_stock
Question → answer

Focused, not firehose

Your agent asks a specific question and gets a specific answer. No HTML soup, no token-burning page dumps — just the fields that matter.

stealthon
renderfirefox
backoff2–15s
captcharetry
Stealth · Firefox · backoff

Built for Amazon's defenses

Stealth mode, Firefox-backed rendering, adaptive 2–15s backoff, CAPTCHA-aware retries, and auto-fallback across direct URL → JS → page hooks.

#productTitlemiss
span.title-textmiss
[data-asin] h1hit
field: title
Multi-selector fallbacks

Self-healing extraction

Multi-selector fallbacks on every field. When Amazon shuffles the DOM, the template tries the next selector instead of returning empty.

json
markdown
hybrid
{ price: "$99.99", in_stock: true }
JSON · Markdown · hybrid

Agent-ready output

Clean Markdown, structured JSON, or a hybrid — whatever format your model handles best. No post-processing layer required.

req 01
req 02
req 03
req 04
Adaptive delays

Respectful by default

Single-stream crawls, adaptive delays, retry-on-captcha. Designed to coexist with Amazon, not fight it.

title
brand
price
rating
reviews
stock
desc
asin
8 fields · ~280ms

Spot-Check Mode

Lightweight lookup grabs title, brand, price, rating, review count, availability, description, and ASIN — perfect for fast verification loops.

Spot-Check Mode

Eight fields. One call. 280 milliseconds.

The fast lookup for confirmation loops — when the agent just needs to confirm "is this still the right product at the right price?" Eight fields. Median 280ms. One round-trip.

field
title
field
brand
field
price
field
rating
field
review_count
field
availability
field
description
field
asin

What teams build with it

From repricing to RAG — one endpoint, many workflows.

Repricing & competitive monitoring

Agents that watch buy-box prices and flag drift the moment it happens.

Catalog enrichment

Fill in missing fields on your own product database from live Amazon listings.

Buyer-side research agents

"Find me the best-reviewed option under $50 with Prime shipping."

Brand protection

Track listing accuracy, A+ content compliance, and seller authorization.

Sourcing & arbitrage

Pull pricing, BSR, and variant data across SKUs in a single pass.

RAG over commerce

Feed clean Amazon data into a vector store the agent can query naturally.

Frequently asked questions

Common questions.

What does Agent Crawl do?
It answers one product question about an Amazon URL or ASIN in a single call, returning the fields your agent asked for: title, brand, price, rating, review count, and availability.
How is this different from running my own scraper?
There is no scraper to babysit. Extraction uses multi-selector fallbacks on every field, so when Amazon shuffles the DOM the template tries the next selector instead of returning empty.
What format does the output come back in?
Structured JSON, clean Markdown, or a hybrid, chosen per request. The point is agent-ready output rather than a page dump your agent has to parse and pay tokens for.
How does it handle Amazon's bot defenses?
Stealth mode, Firefox-backed rendering, adaptive backoff, and CAPTCHA-aware retries, with automatic fallback across direct URL and JS page hooks.
How does this relate to the Selling Partner MCP?
Crawl reads public product pages, so it works on any ASIN including competitors'. The Selling Partner MCP works inside your own seller account. Most teams use both.

The bottom line

The hard part was never fetching the page.

Crawl turns any product page into a structured answer — no DOM wrangling, no retry loops, no anti-bot whack-a-mole. What sets your agents apart is the judgment only your team has. Crawl just clears the way for it.