Home / Docs

Start

AI agent for an online store with automatic stock updates

The templates on this site work with Feedpipe — an app that reads a shop product feed and returns the matching facts when an AI agent asks.

This approach is meant for large catalogs, including 20–50 thousand products. The agent requests what it needs through Function Calling — a product-search tool — instead of loading the whole catalog into context.

What this integration gives you

  • Lower risk of errors and hallucinations. The agent gets product facts from the catalog. That cuts invented specs, prices, and other mistakes, though it does not remove every error.
  • Lower token cost. You do not need to send the full catalog with every request.
  • Easier data updates. Automatic feed refresh and Feedpipe collection reduce manual work.

How current the product data is depends on two stages: feed updates on the shop side, and collection in Feedpipe.

Before you start

You will need:

  • access to the shop settings;
  • a product feed available at a public URL;
  • a Feedpipe account;
  • NEWO and the integration template.

Step 1. Create a product feed

If the shop already has a product feed, use it. If not, set up product export with the CMS tools or a matching plugin.

The feed should include the facts the agent will use for search and answers: titles, prices, stock, descriptions, specs, and product links.

WooCommerce does not publish a catalog address on its own — a feed plugin does. In Woo Product Feed Pro, export Custom / General XML: a public products document with title, price, link, and availability. Paste the address on Connect. The shop adapter is feed_products_xml.

Step 2. Map the fields

Make sure your feed fields match these:

FieldWhat to send
CategoryProduct category
ProductProduct title
SKUProduct SKU
StatusStock status
PriceProduct price
DescriptionProduct description
LinkLink to the product page
ImageLink to the image
SpecsAttributes: size, color, material, and other parameters

Make sure the feed includes everything you want the agent to use. For example, to search by color or size, those attributes must be present in the product data.

Step 3. Set up feed refresh

Turn on automatic feed updates on the shop side.

Pick a frequency based on how often prices, stock, and assortment change. If the CMS can build a fresh feed when the URL is requested, you can use that mode.

After setup, copy the feed URL. Check that it is reachable without logging into the admin panel.

Step 4. Add the feed in Feedpipe

  1. Open Feedpipe.
  2. Paste the feed URL into the field.
  3. Run the check and confirm the data is recognized correctly.
  4. Create the shop and register an account, following the UI prompts.

The first month includes free processing for up to 1 000 products. If the catalog is larger, pick a plan with a matching limit.

Step 5. Confirm email and get the API key

Confirm the email used at sign-up to receive the generated API key.

You need it to connect the template and agent to your shop in Feedpipe. Keep the key secret.

Step 6. Run product collection

Press Collect so Feedpipe can gather data from the feed.

When it finishes, check that products loaded and that titles, prices, stock, and specs look correct.

For ongoing updates, turn on automatic collection in Feedpipe. Feed refresh and collection should run together.

Step 7. Install and connect the NEWO integration

Step-by-step setup is on Connect Feedpipe to NEWO.

Test the agent

Before going live for customers, try a few queries:

  • find a product by title or SKU;
  • pick products from a given category;
  • find products in a price range;
  • check stock and specs;
  • ask for a product that is not in the catalog.

Compare answers with the shop data. If a product or needed attribute is missing, the agent should say so instead of inventing facts.

After a successful check, the integration is ready to use.