BigCommerce AI Chatbot: Add a Shopping Agent

Yes, you can add a BigCommerce chatbot through an app or install an AI assistant’s script through Script Manager. Connect product data and store policies separately, then test the assistant on your storefront. The platform integration guides cover other storefronts. The script displays the chat window. It doesn’t give the assistant permission to read your catalogue or retrieve orders.

BigCommerce Live Chat Apps and AI Assistants

BigCommerce live chat connects shoppers with your staff, while an AI assistant answers automatically from the information you connect. Some BigCommerce apps combine both jobs. Choose by the work you need done, including product recommendations and the route from a recommendation to a purchase.

A marketplace app asks you to authorise its store-data permissions during installation. Its developer maintains the connection and any script the app installs. Review those permissions before approving access. To remove the integration, uninstall the app and inspect Script Manager for any remaining script.

A manually added script puts the assistant on your site without an app installation. Catalogue access is a separate connection, through an API or product feed. You or your agency maintain the script placement and feed configuration, while the assistant provider maintains its service. Remove the manual script in Script Manager to remove the chat window. Retiring catalogue access is a separate step.

shoppa is an AI shopping agent that helps customers find and compare products, then buy in the chat using your catalogue and documents. Its two-line embed puts that conversation on your own site. The catalogue-to-agent process explains the feed and embed stages. Compare live chat software for online stores if your main requirement is a staffed conversation.

Installing Through Script Manager

Add the assistant’s supplied embed through Script Manager, using the placement below for a Stencil storefront. BigCommerce’s Script Manager guide documents the control-panel fields.

1. Open Storefront, then Script Manager.
2. Select Create a Script and give the entry a recognisable name.
3. Choose Footer for Placement and All pages for Location.
4. Choose the Script Type that matches the supplied embed. Paste the complete snippet when the provider supplies script tags.
5. Choose the script category that describes its function, then save.

The categories are Essential, Analytics, Functional and Targeting/Advertising. BigCommerce groups live chat under Functional. With the built-in consent banner enabled, a shopper’s category choice affects whether the script loads. Essential scripts cannot be declined. Don’t use that category simply to make an optional chat window appear before consent.

Keep the provider’s load order intact. For an embed with an initialisation tag after a library tag, changing the first tag to load asynchronously can break that dependency. Record the installed snippet so you can identify and remove it later.

Catalogue Access Through the BigCommerce API

A catalogue connection supplies product records to the assistant. A general approach is to export BigCommerce product data into a JSON or CSV feed, then import it into the assistant’s catalogue. Include variants and their stock, along with product identifiers and prices. Refresh the feed when those values change. A one-off export remains a snapshot until it is replaced.

For a custom API-backed feed, open Settings, then Store-level API accounts and create an account dedicated to the connection. Start with Products set to Read-only. This scope is named store_v2_products_read_only and covers catalogue and price-list reads. Add Store Inventory Read-only (store_inventory_read_only) if the connection reads the Inventory API. Leave unrelated scopes unselected.

BigCommerce’s API account documentation lists these permissions.

Keep management credentials in the feed service, outside the storefront snippet. Reading catalogue data doesn’t require permission to modify products. Ordering and payment workflows need their own integration and permissions. Installing a chat script alone doesn’t establish those workflows.

BigCommerce also has a read-only GraphQL Storefront API for product data. Its guide to product queries explains how a query retrieves prices, options and variations together. Choose the connection that supplies the storefront’s actual product information, including its pricing context.

What the Agent Reads

The product feed gives the assistant the information needed to match a request to products you sell. Store policy documents supply answers about delivery, sizing and returns. Connect both sources before expecting the assistant to answer those questions.

Keep product options attached to the correct variant. A blue shirt in medium needs its own identifier and availability, even when other sizes share the same product page. Preserve required option choices so a customer can select a purchasable combination.

Supply policy pages as supported document imports or URLs. Include delivery regions and exclusions, plus the conditions for returns. Link sizing advice to the relevant product or category. A product description alone cannot explain a store-wide returns rule.

For business-to-business (B2B) stores, price context also matters. BigCommerce price lists can assign variant prices by channel and customer group. A public feed must not expose a customer group’s private prices. Keep signed-in pricing separate from public catalogue answers and test both views.

Multi-Storefront and Checkout Limits

Configure the script and catalogue for each storefront’s channel. In a Multi-Storefront store, find the storefront-specific Script Manager settings through Channel Manager. Each site needs the correct product range and policy documents. A shared control panel doesn’t make every product suitable for every storefront.

BigCommerce’s Scripts API documentation distinguishes storefront placement from checkout placement. All pages includes checkout and order confirmation on supported Stencil stores. Checkout scripts require Optimised One-Page Checkout and theme support for script injection. An API installation that writes checkout scripts also needs Modify Checkout Content permission. That permission is separate from read-only catalogue access.

For an external script on a payment page, follow BigCommerce’s script integrity requirements. Maintain the required integrity hash when the script changes. If the assistant belongs only on shopping pages, choose Storefront pages instead of All pages.

Catalyst has a different boundary. Its script installation guide says it reads All Pages and Storefront Pages scripts, while ignoring other placements such as Order Confirmation. Test the actual checkout separately. For shoppa across several storefronts, contact us to set up the store connections before installing.

Post-Install Tests

Test the assistant on the live storefront after enabling the script. Use an authorised test account for customer-specific checks. Keep payment checks within the store’s approved test-payment process.

• Ask for a product with a specific size and colour. The recommendation must match an available variant, and any basket action must preserve those options.
• Ask about an out-of-stock variant. Compare the answer with the current storefront, then repeat after a catalogue refresh.
• For B2B pricing, compare a guest session with an authorised customer-group session. Each answer must match the price that customer can see.
• Ask about shipping to a region with a restriction. The answer must reflect your published policy without inventing a delivery promise.
• Test order status with an authorised order. Also try an unknown order and another customer’s order. Private order details must remain protected.
• Decline and accept the relevant cookie category in separate sessions. Confirm the chat window follows the consent choice.
• Open product, cart and checkout pages on each storefront. Confirm the script appears only where intended and appears once.

If a check fails, remove the new script while you correct its placement or connection. Keep existing checkout and support paths available. Re-enable the assistant after the same check produces the expected result.