Yandex Commerce Protocol (YCP): Direct Online Store Sales in Alice AIArticle · digest for November 6
Modern editorial vector illustration showing an Alice AI assistant chat and an online store showcase.
Article4 sections≈ 6 min

Yandex Commerce Protocol (YCP): Direct Online Store Sales in Alice AI

You have probably noticed how rapidly everyday online product search is changing. Just a couple of years ago, the buyer's journey looked monotonous: type a query into a search engine, open five tabs of different stores, register all over again on each site, find the right size, and try to navigate a confusing checkout cart. Now that route is breaking down right before our eyes.

Yandex has rolled out an open interaction standard—Yandex Commerce Protocol (YCP). The protocol directly connects the catalogs of independent online stores with the neural assistant Alice AI, product search, and the Ritm recommendation feed. The technology's main goal is to turn conversational AI from a chatty advisor into a full-fledged commerce agent (Agentic Commerce) capable of taking a transaction from the first question to payment without a single redirect to an external website.

At the same time, the store does not turn into a faceless marketplace vendor: the customer base, direct relationship with the buyer, and the order itself go straight into your accounting system, while you control logistics yourself. Let's break down the protocol's internal architecture and how retailers can plug into this new ecosystem.

In the traditional model, search returned so-called "blue links" or product snippets. Even if a search assistant picked the ideal item, the final step was left to the visitor. The person went to an external site, ran into a sluggish mobile interface and mandatory SMS phone verification, and abandoned the purchase. At every intermediate step, conversion inevitably bled away.

YCP eliminates this gap through a seamless agentic model. The buyer chats with Alice in natural language, the system queries connected store servers in real time, verifies stock and pricing, and then builds an embedded purchase card. Payment is charged through the Yandex ID and Yandex Pay ecosystem, while the retailer immediately receives a ready, fully paid order in its CRM or 1C.

In our previous breakdowns on the architecture of conversational AI search, we noted that the future of e-commerce belongs not to passive catalogs, but to end-to-end interaction protocols. YCP has become the first large-scale Russian standard to put this concept into practice.

Five Steps of the Flow: From Chat Query to Delivery

To understand how the mechanics work under the hood, let's trace a typical customer transaction:

  1. Natural Query and Intent Recognition: The buyer no longer needs to type exact technical SKUs. They write in chat using ordinary words: "I need lightweight running shoes for asphalt, size 43, up to 12 thousand rubles with good cushioning." Alice AI analyzes the context of the request, asks one clarifying question if needed (for instance, about surface type or season), and queries the Yandex Goods database.
  2. Initial Matching and Feed Filtering: The algorithm matches parameters against merchant catalogs uploaded via extended product feeds. At this stage, attribute description quality plays a decisive role: if your product card is missing fields for cushioning or sole material, the AI will simply skip your offer.
  3. Synchronous Real-Time Handshake: The moment the user clicks the "Buy" button inside the chat, Alice sends a synchronous API request to your online store's server over YCP. Your server must confirm within fractions of a second:
    • The exact price including active promotions and discounts;
    • Physical availability of the specific size and color in the selected warehouse;
    • Available delivery methods and exact shipping cost to the buyer's address. If the item sold out during the conversation, the protocol honestly informs the user before charging any money, preventing disputes over selling out-of-stock goods.
  4. Universal Checkout: Order placement happens right inside the chat window. The user does not need to re-enter their first name, last name, delivery address, and sixteen bank card digits. Details are pulled automatically from the protected Yandex ID profile, and payment is processed via Yandex Pay, bank cards, or the Split interest-free installment service.
  5. Order Transfer and Fulfillment: After a successful transaction, YCP sends the completed order with a unique number to your system (1C, MoySklad, CRM). Your warehouse packs the parcel and ships it to the customer in your usual way—via your own courier service, CDEK, Russian Post, or Yandex logistics.

Four Ways to Integrate Your Catalog and API

You can connect your store to YCP through the Yandex Goods platform dashboard (merchants.yandex.ru). Depending on your project's technical maturity, four paths are available:

  • Yandex KIT: If your site runs on the Yandex KIT builder, YCP support is enabled in literally one click in the admin panel. It is an out-of-the-box solution for small businesses.
  • Official Module for 1C-Bitrix: The ready-made extension "Yandex Goods: YCP and Feeds" installs from the Bitrix marketplace. The module automatically generates an extended YML catalog, intercepts incoming requests from Alice, checks warehouse stock in real time, and creates the order in the admin panel.
  • Integration via Yandex Market: Stores already working with Market under FBS, FBY, or DBS schemes can enable sales in Alice without separate programming—transactions are simply routed through familiar marketplace dashboards. This trend toward scaling merchant agentic commerce opens access to a new audience without complicating operations.
  • Custom REST API: For stores on non-standard engines (WooCommerce, CS-Cart, OpenCart, or custom frameworks), the dev team implements three endpoints per the YCP specification: cart check, delivery calculation, and order create.

Technical Limits, SLA, and Product Feed Hygiene

For all the technology's surface appeal, YCP imposes strict requirements on the seller's server infrastructure.

The main stumbling block is the server response time SLA. In an AI chat, a person is not willing to wait: if your store's API takes longer than 1.5–2 seconds to respond, the session times out, and Alice politely offers the buyer an alternative from another store. If your 1C database freezes on every synchronous stock query, this sales channel will generate nothing but failures.

The second critical factor is product feed hygiene:

  • Every size, color, and product variant must have a strictly individual unique identifier (offer id / SKU). Grouping shoe sizes under a single shared SKU will trigger a validation error.
  • No marketing clutter in titles. Forget all caps and epithets like "Super discount! Best seller 2026." An item's title must strictly follow the structure: [Product Type] + [Brand] + [Model] + [Key Feature]. That is the exact semantics a neural network reads without error.

YCP is not a replacement for a classic website, but a fundamentally new distribution channel. Those who learn to feed their inventory data to algorithms first will skim the cream off conversational commerce long before this path becomes the universal standard.

Want a deep diveVotes help us choose what to break down in detail. This material has 0 votes.
Sources · 2