Python Amazon Creators API

A Python wrapper for the Amazon Creators API.

Features

• 🎯 Simple object-oriented interface for easy integration
• ⚡ Async/await support for high-performance applications
• 🔍 Product search by keywords, categories, or browse nodes
• 📦 Product details via ASIN or Amazon URL
• 🔄 Item variations support (size, color, etc.)
• 📊 Feeds and reports listing and download URLs
• 💰 OffersV2 support for enhanced pricing and offer details
• 🌍 20 marketplaces supported
• 🛡️ Built-in throttling and retries to avoid API rate limits
• 📝 Full type hints for better IDE support

Table of contents

• Installation
• Credentials
• Quick start
• Usage examples
• Configuration
• Error handling
• Async support
• Working with models
• Documentation
• Contributing

Installation

pip install python-amazon-paapi --upgrade

Python 3.9 or newer is required. Install the async extra to use the asynchronous client:

pip install python-amazon-paapi[async] --upgrade

Credentials

Four values are needed to create a client, and all of them come from the Amazon Associates Creators API portal:

version is the version of the Creators API, not of this library: it is the value Amazon gave you along with the credentials, and it also decides which endpoint issues the OAuth2 token. The accepted values are 2.1, 2.2, 2.3, 3.1, 3.2 and 3.3; any other one raises ValueError when the client is created.

The marketplace is chosen with country, which accepts AU, BE, BR, CA, DE, ES, FR, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US and AE, either as a string or through the Country constants. Pass marketplace instead to give the host directly, such as marketplace="www.amazon.es".

Quick start

from amazon_creatorsapi import AmazonCreatorsApi, Country

# Initialize with your Creators API credentials
api = AmazonCreatorsApi(
    credential_id="your_credential_id",
    credential_secret="your_credential_secret",
    version="2.2",
    tag="your-affiliate-tag",
    country=Country.US,
)

# Get product information by ASIN
items = api.get_items(["B01N5IB20Q"])
print(items[0].item_info.title.display_value)

# Or use Amazon URLs directly
items = api.get_items(["https://www.amazon.com/dp/B01N5IB20Q"])

Every field of a response is optional, as Amazon only sends what it has for an item, so check a value before using it when the item may not carry it.

Usage examples

Get multiple items

items = api.get_items(["B01N5IB20Q", "B01F9G43WU"])
for item in items:
    print(item.images.primary.large.url)

Items come back in the order they were requested, duplicates are asked for only once, and requests with more items than the API accepts at once (10) are split into as many calls as needed, so any amount of items can be requested:

items = api.get_items(asins)  # Any amount of items, split into several calls

Amazon can answer with only some of the requested items, describing the missing ones as partial errors. Those errors are available in the returned list, and unavailable items can be included as an item holding only the ASIN:

items = api.get_items(["B01N5IB20Q", "0000000000"], include_unavailable=True)

for error in items.errors:
    print(error.code, error.message)

for item in items:
    if item.item_info is None:
        print(f"{item.asin} is not available")

Search products

results = api.search_items(keywords="nintendo switch")
for item in results.items:
    print(item.item_info.title.display_value)

A search needs at least one of keywords, actor, artist, author, brand, title, browse_node_id or search_index, and only returns the items available for purchase unless asked otherwise:

from amazon_creatorsapi.models import Availability

results = api.search_items(
    keywords="nintendo switch",
    availability=Availability.INCLUDEOUTOFSTOCK,
)

Get product variations

# Using ASIN
variations = api.get_variations("B01N5IB20Q")

# Or using Amazon URL
variations = api.get_variations("https://www.amazon.com/dp/B01N5IB20Q")

for item in variations.items:
    print(item.detail_page_url)

Get browse node information

nodes = api.get_browse_nodes(["667049031"])
for node in nodes:
    print(node.display_name)

Feeds and reports

Feeds and reports are listed per marketplace, and downloaded through the temporary URL returned by the API:

from amazon_creatorsapi.models import FeedType, ReportType

for feed in api.list_feeds():
    print(feed.feed_name, feed.feed_type, feed.size)

url = api.get_feed("product-feed", feed_type=FeedType.PRODUCT_FEEDS)

for report in api.list_reports():
    print(report.filename, report.report_type, report.last_modified)

url = api.get_report("earnings.csv", report_type=ReportType.CREATOR_CONNECTIONS)

The type is only needed to disambiguate a name available in more than one program, such as a report present in both Creator Central and Creator Connections.

Get the ASIN from URL

from amazon_creatorsapi import get_asin

asin = get_asin("https://www.amazon.com/dp/B01N5IB20Q")

Using OffersV2 resources

items = api.get_items(["B01N5IB20Q"])
item = items[0]
if item.offers_v2 and item.offers_v2.listings:
    listing = item.offers_v2.listings[0]
    print(listing.price.money.amount)
    print(listing.merchant_info.name)

Configuration

Throttling

Throttling value represents the wait time in seconds between API calls, being the default value 1 second. Use it to avoid reaching Amazon request limits.

api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, throttling=4)  # Makes 1 request every 4 seconds
api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, throttling=0)  # No wait time between requests

The interval is kept per client and is safe to share between threads.

Timeout

Timeout value represents the number of seconds to wait for a response before failing. Use None to wait indefinitely.

The default is (5, 25): five seconds to establish the connection and twenty-five to read the response, thirty in total.

api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, timeout=10)  # Fails after 10 seconds
api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, timeout=0.5)  # Fails after half a second

It applies to every API request, including the OAuth2 token refresh.

A pair of (connect, read) seconds bounds each leg of the request on its own. This matters for a host resolving to several addresses: the connect leg is spent once per address, so a single value generous enough to read a slow response is also spent on every address that fails to answer.

# Gives up on an unresponsive address after 1 second, and still reads for 10
api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, timeout=(1, 10))

Retries

Amazon asks clients to back off and try again when it throttles a request or fails to serve it. The client does that on its own, waiting longer before every attempt and honouring the Retry-After header when the API sends it. An expired token is refreshed once and the request is sent again.

api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, retries=5)  # Up to 5 extra attempts
api = AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY, retries=0)  # Fail on the first error

The default is 3 extra attempts, and only the failures that Amazon asks to retry are retried: a rejected request fails right away.

Closing the client

The client keeps a pool of connections open, so it is meant to be created once and reused. Close it, or use it as a context manager, when it is not going to be used again:

with AmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY) as api:
    items = api.get_items(["B01N5IB20Q"])

Custom endpoints

The base URL of the API and the one used to get the OAuth2 token can be replaced, which is useful to run the tests of a project against a mock server. Providing auth_endpoint also makes valid a version that is not in the list yet, so a new one can be used before the library knows about it, as long as it belongs to a family that the library can authenticate: 2.x with Cognito and 3.x with Login with Amazon. A version of any other family is rejected, as a new family brings a new authentication flow and not just another endpoint:

api = AmazonCreatorsApi(
    ID,
    SECRET,
    VERSION,
    TAG,
    COUNTRY,
    host="http://localhost:8080",
    auth_endpoint="http://localhost:8080/token",
)

Error handling

Every error raised by the library inherits from AmazonCreatorsApiError, so a single except covers them all. The message carries the reason given by Amazon, the fields that failed validation and the identifier of the request, which is what Amazon support asks for:

from amazon_creatorsapi.errors import AmazonCreatorsApiError, ItemsNotFoundError

try:
    items = api.get_items(["B01N5IB20Q"])
except ItemsNotFoundError:
    print("The item is not available")
except AmazonCreatorsApiError as error:
    print(error)

Async support

For async/await applications, use the async version of the API with httpx:

pip install python-amazon-paapi[async] --upgrade

The async API provides the same methods, parameters and errors as the synchronous version, and they are called with await:

from amazon_creatorsapi import Country
from amazon_creatorsapi.aio import AsyncAmazonCreatorsApi

# Use as async context manager (recommended for connection pooling)
async with AsyncAmazonCreatorsApi(
    credential_id="your_credential_id",
    credential_secret="your_credential_secret",
    version="2.2",
    tag="your-affiliate-tag",
    country=Country.US,
) as api:
    items = await api.get_items(["B01N5IB20Q"])
    results = await api.search_items(keywords="laptop")
    variations = await api.get_variations("B01N5IB20Q")
    nodes = await api.get_browse_nodes(["667049031"])
    feeds = await api.list_feeds()
    reports = await api.list_reports()

# Or use without context manager (creates new connection per request)
api = AsyncAmazonCreatorsApi(ID, SECRET, VERSION, TAG, COUNTRY)
items = await api.get_items(["B01N5IB20Q"])

Note: outside async with, every request opens and closes its own connection, so there is nothing to release and the async client has no close method. Use async with when making more than one call, to keep the connection open between them.

Working with models

All SDK models are re-exported through amazon_creatorsapi.models for convenient access:

from amazon_creatorsapi.models import (
    Condition,
    GetItemsResource,
    Item,
    SearchItemsResource,
    SortBy,
)

# Use Condition enum for filtering
items = api.get_items(["B01N5IB20Q"], condition=Condition.NEW)

# Use SortBy enum for search ordering
results = api.search_items(
    keywords="laptop",
    sort_by=SortBy.PRICE_COLON_LOW_TO_HIGH,
)

# Specify which resources to retrieve
resources = [
    GetItemsResource.ITEM_INFO_DOT_TITLE,
    GetItemsResource.OFFERS_V2_DOT_LISTINGS_DOT_PRICE,
]
items = api.get_items(["B01N5IB20Q"], resources=resources)

Every method asks for all the resources of its operation when resources is not given. Narrowing the list makes the response smaller and faster, and the fields left out come back as None.

Documentation

• 📖 Full documentation
• 📘 Usage guide
• 🔀 Migration guide from version 6
• 🔀 Migration guide from amazon_paapi
• 📋 Changelog
• 💬 Telegram support group

Contributing

Contributions are welcome. See CONTRIBUTING.md for the full guide, or get started with:

git clone https://github.com/sergioteula/python-amazon-paapi.git
cd python-amazon-paapi
uv sync --extra async
make setup
make test

Pre-commit hooks run Ruff, mypy and the tests before each commit.

License

MIT License © 2026 Sergio Abad