Get started
Muse and Grok Bot are Relay’s first integration targets.
The initial connector is being designed around the Model Context Protocol (MCP), so an assistant can use Relay’s commerce tools during a conversation.
For integration inquiries and availability updates, email info@askrelay.io. Tell us which assistant or agent platform you are building for and what you would like Relay to help users find.
Muse and Grok Bot first
Relay’s initial focus is integration with Muse and Grok Bot.
The goal is straightforward: let people search, compare and discover products through the assistant they already use.
Additional assistants and agent platforms may be supported over time.
What Relay does
Relay connects AI assistants to live commerce.
A single Relay request can search and resolve products across participating retailers, marketplaces, commerce platforms, affiliate networks, and product-data sources.
Relay normalizes the results into a consistent commerce model so the assistant can search, compare, select, and continue toward purchase without needing to understand each underlying commerce system individually.
One connector. Multiple commerce sources.
Relay is designed to act as a single commerce connection for an AI assistant.
Behind that connection, Relay can search and orchestrate across multiple product catalogs, retailers, marketplaces, commerce platforms, affiliate networks, and product-data sources.
The assistant does not need to integrate separately with each underlying commerce source.
Relay handles:
- Source selection
- Product discovery
- Product normalization
- Cross-source comparison
- Offer resolution
- Merchant selection
- Attribution
- Purchase handoff
Different commerce sources may provide different capabilities, inventory, pricing, availability, or checkout paths.
Relay brings those sources together through one consistent interface so the AI assistant can focus on the user's intent rather than the underlying commerce infrastructure.
Commerce sources
Relay is designed around a source-adapter architecture.
Different commerce sources may provide different capabilities, including:
- Product discovery
- Live catalog data
- Pricing
- Availability
- Product variants
- Seller or merchant information
- Historical pricing
- Merchant offers
- Cart creation
- Checkout preparation
- Affiliate attribution
Relay can combine information from multiple sources while presenting a consistent interface to the AI assistant.
The availability and freshness of specific fields depend on the underlying commerce source.
Initial capabilities
The initial connector is designed around six core capabilities. The descriptions below explain their intended behavior; final input and output schemas will be published with the integration documentation.
search_products
Find relevant products across available commerce sources using natural-language intent and optional structured constraints.
Typical inputs may include:
- Search query
- Product category
- Brand
- Minimum and maximum price
- Currency
- Product attributes
- Condition
- Availability requirements
- Delivery region
- Other user-supplied constraints
Relay may query multiple commerce sources and normalize the resulting products before ranking and returning them.
get_product
Retrieve current information about a specific product or product family.
Where supported, Relay may resolve equivalent products or offers across multiple merchants.
compare_products
Compare products using available information such as:
- Product attributes
- Current price
- Availability
- Condition
- Merchant
- Shipping
- Offer information
- Product variants
find_similar
Find alternatives or substitutes based on the selected product or commerce intent.
This can include:
- Lower-priced alternatives
- Comparable brands
- Similar products
- Different variants
- Replacement products
- Available alternatives to an out-of-stock item
get_offers
Retrieve available merchant offers for a selected product where Relay can identify multiple relevant sellers or commerce paths.
An offer may include:
- Merchant
- Current price
- Currency
- Availability
- Product variant
- Condition
- Shipping information
- Purchase URL
- Attribution information
- Offer freshness
get_purchase_link
get_purchase_link prepares an agent-execution purchase handoff.
It does more than return a destination URL.
Relay may return:
- Selected product
- Selected offer
- Merchant
- Product variant
- Size, color, condition, or other selected options
- Quantity
- Expected price and currency
- Availability
- Shipping information when known
- Affiliate or attribution metadata
- Current merchant destination
- Offer freshness timestamp
- Available cart or checkout context
- Machine-readable instructions for the next purchase step
The AI assistant can then continue the purchase using its own browser, computer environment, or supported commerce protocol, subject to the assistant's own user-approval and security requirements.
Relay does not require payment-card credentials to prepare a purchase handoff.
Purchase paths
Relay may support more than one purchase path depending on the underlying merchant or commerce platform.
Merchant browser handoff
Relay returns an attributed merchant destination and the structured context the assistant needs to continue checkout on the merchant's website.
Agent-compatible cart or checkout
Where an underlying commerce platform supports agent-compatible cart or checkout capabilities, Relay may prepare the appropriate cart, checkout session, or continuation path for the assistant.
The applicable merchant remains responsible for the underlying commerce transaction unless explicitly stated otherwise.
Attribution and purchase routing
Product discovery and purchase attribution are separate functions within Relay.
A product may be discovered through one commerce source while its purchase is routed through another eligible commerce relationship.
Before preparing a purchase handoff, Relay may resolve the selected product or offer through an available affiliate, referral, marketplace, or merchant attribution path.
This allows Relay to preserve commercial attribution without allowing affiliate compensation to determine organic product ranking.
When an attributed purchase path is available, Relay preserves the required referral or attribution information through the purchase handoff.
Not every product, merchant, or transaction is eligible for commission. Where no eligible monetization path exists, Relay may still return the product if it is relevant to the user's request.
How Relay works
Intent
The AI assistant sends Relay the commerce request and relevant user constraints.
Source selection
Relay determines which available commerce sources are appropriate for the request.
Search
Relay queries one or more live product catalogs, marketplaces, merchant sources, or product-data providers.
Normalize
Relay converts results into a common product and offer model.
Policy
Results pass through Relay's commerce-policy and eligibility layer.
Rank
Eligible products are ranked according to their usefulness to the request.
Resolve
Relay identifies the selected product, variant, merchant, and current offer.
Purchase handoff
Relay prepares the appropriate destination, attribution data, and purchase context so the AI assistant can continue toward checkout.
Outcome
Where available, Relay records referral, transaction, or conversion signals for attribution, analytics, and aggregated commerce intelligence.
Ranking and promoted placements
Relay distinguishes organic product ranking from commercial placement.
Organic ranking may consider:
- Semantic relevance
- User constraints
- Price
- Availability
- Product attributes
- Condition
- Merchant quality
- Shipping
- Current offer information
Affiliate commission does not determine Relay's organic relevance ranking.
Some underlying commerce platforms may make separately identified promoted or sponsored product placements available.
Where Relay chooses to surface such placements, they should be distinguishable from organic results.
Commerce Intent Graph
Relay may convert each commerce request into a structured intent event.
Signals may include:
- Product category
- Brand
- Product or model
- Desired attributes
- Price range
- Condition
- Availability requirements
- Purchase urgency
- Coarse geographic region
- Products or brands considered
- Search outcome
- Merchant interaction
- Purchase handoff
- Conversion attribution where available
These events can contribute to Relay's:
Demand Graph What consumers are actively trying to find.
Consideration Graph Which products, brands, and alternatives compete for the same intent.
Conversion Graph How product intent progresses toward merchant interaction and purchase.
Relay may also identify unmet demand, including searches where suitable inventory or offers cannot be found.
Aggregated and de-identified insights may be used to develop market-level commerce intelligence.
Relay does not sell identifiable user-level intent profiles.
Product and offer information
A product describes the item. An offer describes how a particular merchant is selling it, including available price, condition, stock and delivery details.
Relay brings these details together so the assistant can help users compare options across participating sources.
Catalog coverage and the details available for each product depend on the connected sources. Prices, inventory and shipping information can change. Check the merchant’s page for the latest details before purchasing.
Questions and support
Where do users complete a purchase?
With the participating retailer or marketplace. Relay returns a structured purchase handoff, and an AI assistant may use its own browser or computer environment to continue checkout on the user's behalf, subject to its own approval and security requirements.
Who can help with an existing order?
Contact the merchant where you completed the purchase for help with delivery, returns, refunds or warranties.
How can merchants work with Relay?
Visit For Merchants or email info@askrelay.io to discuss your catalog and potential integration.
How can developers get in touch?
Email info@askrelay.io for product questions, integration inquiries and availability updates.