Helium 10 MCP with Claude: Amazon data plus AI answer visibility

The Helium 10 connector lets Claude read your Helium 10 account: keywords, listings, rank history, advertising, inventory and profit. Its Agent Index listing shows 34 tools, 33 of them reads, and names analyze_keywords, get_keywords_by_asin, get_keywords_rank_history, get_listing_score and compare_listings among them. Helium 10's own guide says it needs a Diamond plan or higher and covers Amazon marketplaces from the US to Mexico, plus Walmart and TikTok Shop.

That is the marketplace half. It describes how shoppers search inside Amazon. It does not describe what an answer engine tells a shopper who never opens Amazon. Linkeddit is the other half: Answer Radar records which brands the five engines name and recommend for your buying questions, and Compete tracks competitor moves. Both come with one plan at $60 a month. Linkeddit writes no listing copy and changes nothing in your account, so Claude does the analysis.

One boundary matters. Linkeddit measures ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. It does not measure Amazon's own search or its shopping assistant, and Helium 10's own tools are the place for those.

Start from the questions, not the page. A shopper asking ChatGPT what to buy gets a short list of products, and the first job is to learn which questions skip you. Linkeddit lists those gaps: the buying questions where an engine recommended somebody else or nobody, with the domains it cited. Its content view then scores each tracked page against each measured question and names the requirements the engines agreed on that the page misses. It reports what is missing and writes no advice.

The rest is a short loop. Claude reads the gap list from Linkeddit, pulls the matching products from Helium 10 with search_products and get_listing_score, and compares your listing text with the requirements. Helium 10 is where the product truth lives: the attributes, the keywords you rank for and the competitor ASINs. Claude then drafts changes for the pages you own, and you approve each one. Helium 10's connector does not publish to your store, so the edit happens wherever the page lives. If that is Shopify, use the Shopify version at /integrations/shopify/product-pages-chatgpt-shopping, where the Shopify connector can make the edit.

Independent guides agree on what the engines read. Precis, an agency, reports that ChatGPT's product carousels are built from the top 40 organic Google Shopping results, re-ranked by an AI layer, based on Peec AI research of more than one million shopping queries, and that products rated below three stars are typically filtered out. WorkDuo, a competing tracker, lists missing attributes, inconsistent details across sources and near-duplicate thin pages as the common mistakes. Treat these as vendor claims and check them against your own measured gaps.

Amazon sellers add a second surface. Helium 10's AI Listing Builder post says a listing now serves keyword matching, Amazon's Rufus assistant and intent-driven discovery at once, and that Rufus indexes up to the tenth bullet point. Helium 10 covers that surface. Linkeddit covers the web side: your own product pages, brand pages and the third-party pages that answer engines cite. A seller with both needs both.

An ecommerce SEO on Reddit described the manual version: "Right now I’m checking a fixed list of buying questions every month, noting whether an AI Overview appears and who it cites, and pairing that with branded search and GA4 referrals from AI platforms." Linkeddit replaces the checking with measured samples, and keeps the fixed list.

Claude can use Helium 10 to find the words to fix it with, and nothing more. On Linkeddit's test project, the top open gap reads "Not represented for: 'Umami vs Matomo for self-hosted analytics'". Plausible Analytics was not named in any of 3 successful answer samples, while docs.umami.is was cited 5 times, umami.is 4 times and v2.umami.is 3 times. That is the shape of a gap record: who won, and which pages the engines cited.

For an Amazon brand the same record would show, for a question like best waterproof phone pouch, which domains the engines cited. Claude then asks Helium 10 which keywords and attributes the competing listings rank for through get_keywords_by_asin, and whether your listing states them. Helium 10 describes the connector as mostly read-only, with selective write functions for tracked keywords, costs and ad campaigns, and says it cannot write to Seller Central directly, so publishing the fix stays a step you do yourself.

Partly, and the two connectors cover different competitors. Helium 10's Market Tracker updates its data once a day and shows each tracked product's reviews, rating, price, estimated sales and revenue, best sellers rank and market share. Helium 10's guide says Diamond allows 5 markets and recommends 100 to 150 ASINs per market. Through the connector, get_competitor_overview and list_tracked_competitor_keywords bring that into Claude.

Linkeddit watches the competitors that sit outside Amazon: pricing pages, changelogs and blogs, each signal graded by priority with a reason. On the test project the latest week, 2026-10-26, held 4 signals, all low priority, including two Umami blog posts about its v3 release and roadmap. A brand that sells on Amazon and on its own store can ask Claude to hold both lists together. Linkeddit is not a repricing tool, and Helium 10 is the place to track an ASIN's price.

Yes, and Helium 10 itself leads on this. Its case study of the agency YourSeller says a team managing four brands and about 100 SKUs moved from running Cerebro, Magnet, Black Box and Market Tracker one at a time to a single prompt in Claude, and reports one waterproof pouch moving from rank 20 plus to rank 6 in its bestseller category. These are Helium 10's own figures.

Linkeddit adds the question before the keyword: which buying questions the answer engines are asked in your category and who they name. A keyword with search volume on Amazon can sit beside a question where no engine recommends you. Claude can rank the two lists together and draft a page outline for the overlap, and you decide what is worth publishing. Helium 10 also offers a free Keyword Opportunity Finder skill for Claude that reads a Cerebro export, which needs no connector at all.

Helium 10's guide includes 1,000 API calls a month with a Diamond plan, shared with any sub-users. A simple question uses 1 to 2 calls and a comprehensive report 5 to 10. Each connection reads one account, so a multi-brand agency connects each account separately. Responses take 5 to 15 seconds, and the data refreshes periodically rather than in real time.

Plan the prompt around that budget. Pull the Linkeddit gap list first, which does not use Helium 10 calls, then ask Helium 10 about the top three products only. Check any number Claude quotes against the Helium 10 dashboard before you act on it, and keep the prompt to the products the gaps point to.