Back to glossary

Definition

Merchant Agent

An AI agent operating on the merchant's side of commerce — selling to shoppers and their buyer agents, and in enterprise usage, automating merchant operations like merchandising and campaigns.

Definition

A merchant agent is an AI agent that acts for the merchant in agentic commerce. The term covers two related things. Customer-facing: an agent that sells — answering shopper questions, guiding choices, transacting with outside buyer agents over protocols like the Universal Commerce Protocol. Operational: an agent that runs merchant workflows — merchandising decisions, promotional campaigns, inventory and pricing moves. Enterprise platforms lean on the operational sense; Salesforce's Agentforce Commerce, for example, ships a product named Merchant Agent (generally available July 2026) aimed at retail operators.

Why it matters

Commerce is heading toward agent-to-agent transactions: a shopper's buyer agent discovers a store, reads its manifest, negotiates capabilities, and completes a purchase — machine to machine. In that world the merchant needs a capable party on its own side: something that can answer correctly, represent the catalog deeply, honor policies exactly, and close. Merchants that field a real agent convert the demand that buyer agents route; merchants that don't are represented by whatever the buyer agent could scrape.

Merchant agent vs brand agent

Same side of the table, different center of gravity. "Merchant agent" frames the seller as an operator — a business with a catalog, campaigns, and operations to automate — and is the vocabulary large retail platforms use. "Brand agent" frames the seller as a brand — an identity with a voice, whose agent must sell the way the brand would. For a D2C brand the distinction is practical: operational automation saves cost, but the revenue question is whether the agent representing you to shoppers and buyer agents actually sells.

How one gets deployed

In practice a merchant agent is built on the merchant's real data: the catalog with its variants and constraints, policy documents, past customer conversations, and brand guidelines. It then fronts two surfaces — the merchant's own storefront, where it converses with human shoppers, and the machine-readable layer (structured feeds, UCP manifests, agent endpoints) where it answers other agents. The quality bar is groundedness: a merchant agent that invents an answer about shipping or stock does more damage than no agent at all.

See merchant agent in action.

Book a call to talk through your store — live on your catalog the same day.