Product in development by SENDER Inc. - Sender Seller Agent
A sales agent that answers AI shopping agents on behalf of online stores.
When an AI shopping agent asks a store for "noise-cancelling earbuds under 20,000 yen that arrive tomorrow", Sender Seller Agent answers on behalf of the store. It returns accurate product terms, offers only the prices and bundles that the store has approved, and hands the order to checkout. It does not provide a chat interface for human shoppers.
Why we are building it
Purchases made by AI agents on behalf of people became a standard feature of major platforms in 2026. AI shopping agents do not read product pages or advertisements. They compare structured terms and confirmed values. Most online stores are not prepared to answer them correctly.
- January 2026
- Google and Shopify published the Universal Commerce Protocol (UCP). OpenAI and Stripe operate the Agentic Commerce Protocol (ACP), and Microsoft launched Copilot Checkout. AI agents can now search products, build carts and complete checkout through shared standards.
- January 2026
- Shopify reported that orders originating from AI tools grew about 11 times between January 2025 and January 2026.
- March 2026
- Shopify enabled Agentic Storefronts for all merchants in the United States.
- July 2026
- In Japan, about 20% of the order value on Yahoo! Shopping came through AI agents, according to LY Corporation.
Sources: announcements by Shopify, Google, OpenAI, Microsoft and LY Corporation, and press reports, as of October 2026. These figures are public information and are not results of SENDER Inc.
What it does
01
Product search and comparison
Selects products that match the request of the AI shopping agent, and returns structured reasons why each product matches or does not match. When no product matches, it returns "no match" and the conditions under which a product would match.
02
Confirmed sales terms
Returns the current price, stock, delivery date, warranty and return policy from the store data, with the time of retrieval. The language model does not generate these values.
03
Offers and quotes
Combines bundles, accessories and delivery options within the range the store has approved, and issues a quote with an expiry time. The system enforces the maximum discount and the minimum margin set by the store.
04
Hand-off to checkout
Creates a cart and checkout in the format of the protocol used by the AI shopping agent, and returns order and delivery status only to parties with permission.
An example exchange
AI SHOPPING AGENT
I am looking for noise-cancelling earbuds for commuting, under 20,000 yen, delivered tomorrow.
SENDER SELLER AGENT
Two products match. The recommended product is NC-Pro 2: 18,800 yen including tax, 12 units in stock as of 11:20, delivered tomorrow for orders placed by 14:00.
AI SHOPPING AGENT
Can you add the extended warranty and keep the total under 20,000 yen?
SENDER SELLER AGENT
The store does not discount this product. With the approved warranty bundle, the total is 19,900 yen. This quote expires in 30 minutes. A checkout link has been issued.
Product names and prices in this example are fictional.
What the AI does, and what the system enforces
- Handled by the language model
- Interpreting the request written in natural language, and writing the explanation that accompanies each answer.
- Enforced by the system
- Prices, stock, delivery dates, discount limits, minimum margins, approved bundles, quote expiry, and order processing. These follow the rules set by the store.
- Handed to a person
- Invoice payment, large orders and return exceptions are sent to the store staff with a decision ID.
Requests from AI shopping agents are treated as requests, not as instructions to the system. Cost prices and internal estimates are never returned to outside agents. Every query, quote, order and return is recorded under one decision ID, so the store can check what was offered, to which agent, and on what basis.
How it connects
Step 1
AI shopping agent
An AI assistant or a purchasing agent sends a request on behalf of a consumer or a company.
Step 2
Connection layer
Adapters for a common REST API, UCP, ACP and MCP absorb the differences between protocols.
Step 3
Authentication and limits
Permissions and request limits are set for each connected agent.
Step 4
Sender Seller Agent
Interprets the request, selects products and proposes an offer.
Step 5
Sales rule engine
Applies price, stock, minimum margin, quote and checkout rules.
Step 6
Store systems
Product, stock and order data from Shopify and other e-commerce platforms.
Built with Claude
We are developing Sender Seller Agent on Claude, the language model developed by Anthropic. In our design, Claude interprets requests from AI shopping agents, calls tools that read product terms and apply sales rules, and writes the explanation in each answer. Prices, stock and order processing are handled by deterministic rules, not by the model.
Claude is a trademark of Anthropic, PBC. This page does not indicate a partnership with or endorsement by Anthropic.
Current status
- Now (October 2026)
- Specification research and prototype development. The product is not yet commercially available.
- Next (planned for 2027)
- A connection pilot with pilot stores and AI agent providers. We will measure query volume, quote accuracy, order completion rate, store margin and inference cost.
- After the pilot
- A Shopify app for stores, starting with the United States.
Partners we are looking for
Pilot stores
Online stores on Shopify or their own platforms that sell products with many options, such as consumer electronics and accessories, sporting goods and specialty foods, and that compete on bundles, warranties and delivery terms.
AI agent providers
Companies that provide AI shopping assistants, purchasing agents or procurement agents, and that need confirmed terms and quotes from many stores through one connection.