--- title: "Amazon Creators API Integration" description: "The Amazon Creators API is the live successor to Product Advertising API 5.0 for Amazon Associates. Agents search the Amazon catalog, pull titles, prices, images and availability for ASINs, and watch listings for price and stock changes." url: https://flowrunner.ai/integrations/amazon-pa-api-v5 date_modified: 2026-09-04T19:08:45-07:00 --- # Amazon Creators API [E-commerce](https://flowrunner.ai/integrations/category/e-commerce) The Amazon Creators API is the live successor to Product Advertising API 5.0 for Amazon Associates. Agents search the Amazon catalog, pull titles, prices, images and availability for ASINs, and watch listings for price and stock changes. 6 actions · OAuth · available [Amazon Creators API website](https://creatorsapi.amazon/) · [Platform Documentation](https://affiliate-program.amazon.com/creatorsapi/docs/en-us/paapiv5-deprecation) · Capability data verified 2026-08-26 1. A scheduled refresh of the watched ASIN list begins 2. The agent reads the ASINs from the sheet or table that holds the watchlist 3. Get Items looks up ten ASINs in a single call rather than looping one at a time 4. Get Item Offer flattens price, availability and the affiliate link into one row per product 5. Each row is compared against the value recorded on the previous run 6. Movers post to the merchandising channel with the old price, the new price and the stock flag 7. An editor confirms which changes are worth republishing before any content is touched ## What This Integration Enables This is not a storefront connector. The Amazon Creators API is a metered, read only window onto somebody else's catalog, and every honest workflow built on it starts by admitting that. FlowRunner agents search a marketplace by keyword, brand, author or browse node, look up as many as ten ASINs in a single transaction, resolve a product's whole variation family from any one member of it, and pull a flattened affiliate row carrying title, image, price, currency, availability and the purchase link. What agents cannot do is write anything, and what they must not do is pretend a search saw more than it did. That second point is the discipline the connector is really about. A single search can never reach more than one hundred items, while the response happily reports a total in the hundreds of thousands. An agent that treats that total as the size of its evidence will report a confident conclusion drawn from a sliver of the catalog. FlowRunner's answer is [human-in-the-loop](https://flowrunner.ai/concepts/human-in-the-loop) applied to the reading step rather than the writing step: agents do the sweeps, the counting and the comparison, and a person sees the gap between what was reported and what was reachable before anything gets published on the strength of it. ### Without FlowRunner **Prices quietly age**: Published roundups keep quoting a figure the marketplace moved on from weeks ago **Research by browser tab**: Checking availability and finding every size of a product means opening product pages by hand **Budget spent without a plan**: A per product loop burns the daily transaction allowance long before the watchlist is covered ### With FlowRunner **Prices refresh on a schedule**: Titles, images, prices and stock flags are pulled on a cadence the team sets **Variation families resolved in one call**: One child ASIN is enough to recover every color and size in the family **Budget spent deliberately**: Ten ASINs per transaction, with the sweep reporting what it could not reach ## Use Case Scenarios ### A watchlist that refreshes itself A list of ASINs lives in [Google Sheets](https://flowrunner.ai/integrations/google-sheets), one row per product the team writes about. On a schedule the agent reads the rows, batches them ten at a time into Get Items, and calls Get Item Offer for the flattened price, currency, availability and affiliate link. Rows are matched back on the asin field rather than on request order, because the response does not preserve it and invalid ASINs move to a separate errors list. Movers append back to the sheet with a timestamp, and a summary posts to [Slack](https://flowrunner.ai/integrations/slack) naming only the products whose price or stock state actually changed. ### Catalog enrichment without the copy and paste An internal product catalog in [Airtable](https://flowrunner.ai/integrations/airtable) carries names and internal codes but no imagery and no competitive price. The agent runs Search Items against the marketplace for each record, uses Get Variations to pull the full family when the match is a single variant, and writes the title, image URL and price back onto the record. Get Browse Nodes resolves the category tree so a search can be narrowed to the niche the product actually sits in rather than the whole marketplace, which is usually the difference between a useful match and a plausible wrong one. ### The sweep that reports its own blind spot A merchandising analyst wants the newest arrivals in a category. The agent calls Search Items All Pages, which walks consecutive pages and pauses between them so the account's transaction rate is respected, and stops at the first empty page. It then reports two numbers side by side: the total the marketplace claims for that query, and the number of items the ceiling actually allowed it to retrieve. The analyst reads both before deciding whether the shortlist is a fair sample or a coincidence of ranking, and the workflow never writes a recommendation the evidence cannot carry. ## Human-in-Loop Highlight The costly operation here is not a write, because there are none. It is Search Items All Pages spending a finite daily transaction budget to produce an answer with a hard ceiling under it. So the sweep agent stops before the finding is used. It posts: "This query reports a total in the hundreds of thousands. The ceiling allowed 100 of them, retrieved across ten pages and ten transactions against today's allowance. The three cheapest results are attached. Do you want this treated as a price floor, or should I narrow it with a browse node first?" The analyst answers, and the difference matters: narrowing by browse node makes the same 100 items a defensible sample of a niche instead of an arbitrary slice of a category. The agent did the walking and the arithmetic. The judgment about whether the evidence supports the claim stayed with the person who has to stand behind it. Agent processes routinely Detects exception requiring judgment Clear match Continues automatically Ambiguous Routes to human via preferred channel Human decides Agent resumes with decision ## Agent Capabilities 6 actions ### Search 2 - **Search Items** Searches a marketplace by keywords, title, brand, author, artist, actor or browse node and returns matching products with the fields the workflow asks for. At least one search term is required, and a query that matches nothing returns an empty list rather than an error. - **Search Items All Pages** Runs the search across consecutive pages and concatenates the results, stopping at the first empty page and pausing between pages so the account's transaction rate is respected. Use it when a workflow wants the widest reachable slice in one step, and read the ceiling caveat before trusting the count. ### Catalog Lookup 3 - **Get Items** Looks up between one and ten ASINs in a single call and returns the requested fields for each. Batching is the whole point, because one call is one transaction no matter how many ASINs it carries. - **Get Variations** Returns the child variations of a parent ASIN, the dimensions they vary on, and the lowest and highest price across the family. Passing any child still resolves the whole family, which is how a workflow finds every color of a product it only knows one version of. - **Get Browse Nodes** Resolves up to ten browse node identifiers into their names, ancestors and children. This is how an agent walks the category tree to find the node that narrows a search to a real niche. ### Affiliate Rows 1 - **Get Item Offer** Looks up up to ten ASINs with title, large image, price, availability and buy box fields pinned, and flattens each into a single row with the affiliate link included. This is the operation most affiliate workflows want, because it removes any need to reason about the field selection model, and it keeps the untouched original listing alongside each row. ## Frequently Asked Questions ### What can FlowRunner do with Amazon Creators API? FlowRunner agents can run Search Items, Get Items, and Get Variations in Amazon Creators API, plus 3 more actions. ### Does connecting Amazon Creators API to FlowRunner require OAuth? Yes. Amazon Creators API connects to FlowRunner with OAuth 2.0, so agents authenticate without handling raw credentials. ### Can Amazon Creators API trigger a FlowRunner workflow automatically? Amazon Creators API doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger. **Work at Amazon Creators API?** This integration exposes Amazon Creators API to AI agents on every FlowRunner plan, including through MCP, at no cost to you. [See what FlowRunner offers integration partners](https://flowrunner.ai/integrations/partners), including how to keep this page current. --- Markdown version of https://flowrunner.ai/integrations/amazon-pa-api-v5. Site index: https://flowrunner.ai/llms.txt