Agentic commerce glossary

Agent experience (AX) is how easily AI agents can use a product or website on their own: find it, understand it, take actions and recover from errors. Netlify coined the term in 2025 and defines it as the holistic experience AI agents have as the user of a product or platform.

Example: Prices an agent can read without logging in, and an API it can call, improve a store’s agent experience.

Read more: How to prepare your site for AI shopping agents

Agent Payments Protocol (AP2) is an open protocol from Google for payments that AI agents make on a user’s behalf. It uses signed mandates: a checkout mandate records what the agent and the merchant agreed, and a payment mandate authorizes the payment. AP2 works as an extension of A2A, MCP and UCP.

Read more: How the commerce protocols compare

Agentic checkout is a purchase that an AI agent completes for its user, after the user confirms, without the user filling in the merchant’s checkout page. The merchant still takes the order and the payment. Examples include OpenAI’s Agentic Checkout Spec for ChatGPT and Google’s agentic checkout in Search.

Example: A shopper asks their assistant to reorder running shoes, approves the total, and the assistant places the order with the store.

Read more: How the commerce protocols compare

Agentic commerce is buying and selling in which AI agents do part of the work: they search, compare, ask questions and complete purchases for the people or businesses they act for, with their approval. On the selling side, a business can run its own agent to answer those agents and make offers.

Example: A customer asks their assistant for a coffee grinder under $200. It compares stores, asks two questions and brings back one choice to approve.

Read more: What is agentic commerce?

Agentic Commerce Protocol (ACP) is an open standard, developed by Stripe and OpenAI, for checkout between buyers, AI agents and businesses. A business exposes its checkout to agents and stays the merchant of record. ChatGPT was the first AI platform to use it. ACP is open source under the Apache 2.0 license.

Read more: How the commerce protocols compare

An agentic storefront is a store’s presence inside AI assistants: its products, prices and checkout made available so a shopper can find and buy them in the conversation instead of on the store’s website. Shopify’s Agentic Storefronts feature does this for its merchants in AI channels such as ChatGPT, Gemini and Microsoft Copilot.

Example: An agentic storefront gets a store’s products into the assistant’s answer. Winning the comparison that follows is a separate job.

Read more: How to sell to AI shopping agents

A budget, in haggl.ai, is the total amount a merchant sets aside for the discounts its sales agent can give, with haggl.ai’s fee included. The merchant caps it and can pause new offers at any time. Every offer stays within what is left of the budget and within the max discount for one shopper.

Example: You cap the budget at $1,000, and every discount your sales agent gives, plus our fee on it, comes out of it.

Read more: How much discount should you authorize?

Business-to-agent (B2A) is a model in which a business sells to its customers through their AI agents, so it designs its products, data and checkout for software agents as well as for people. Kantar, the research firm, describes it as companies building for software agents first and humans second.

Read more: Our conviction: personal agents and commerce

A buyer’s agent is an AI agent that buys for a business or a person: it compares plans, terms and offers, then brings its user a choice to approve. haggl.ai uses the term for software and financial services, where the buyer is often a team. In retail, haggl.ai calls the same role a shopper’s agent.

Example: A buyer’s agent asks a software company’s sales agent which plan fits a 40-person team that renews in March.

Read more: Agentic commerce for SaaS

llms.txt is a proposed standard for a Markdown file, usually at a site’s root (/llms.txt), that gives AI agents and language models a short summary of the site and links to its most useful pages. Jeremy Howard proposed it in September 2024, and the proposal is published at llmstxt.org.

Example: haggl.ai publishes one so agents can learn what haggl.ai does and which pages to read.

Read more: haggl.ai’s llms.txt

Max discount, in haggl.ai, is the most that any one shopper can get from a merchant’s sales agent, set by the merchant as a percentage of the price. The sales agent can offer less but never more. It works alongside the budget, which caps the total the sales agent can give across all shoppers.

Example: With a 15% max discount on a $200 jacket, no shopper’s agent can be offered more than $30 off.

Read more: How much discount should you authorize?

Model Context Protocol (MCP) is an open-source standard for connecting AI applications, such as Claude and ChatGPT, to outside tools, data and workflows. Anthropic introduced it in 2024 and in December 2025 donated it to the Agentic AI Foundation, part of the Linux Foundation. A business can run an MCP server so agents can call its services.

Example: haggl.ai runs MCP servers: one where shoppers’ agents ask for offers, and one where merchants run their campaigns from Claude, ChatGPT or Gemini.

Read more: The haggl.ai protocol · Install with your coding agent

Performance marketing for agentic commerce is winning customers who shop through AI agents and paying only for results. A merchant’s sales agent makes each shopper’s agent a personalized offer within a budget the merchant caps, and the merchant pays when the shopper buys. It is the category haggl.ai works in.

Read more: Pay-for-performance pricing · What haggl.ai does

A personal agent is an AI assistant that acts for one person: with their permission, it uses what it knows about them to do tasks such as finding, comparing and buying products. ChatGPT, Claude and Gemini can act as personal agents. Inside a purchase, haggl.ai calls it the shopper’s agent.

Example: People now shop with personal agents: one request to an assistant replaces a dozen browser tabs.

Read more: Our conviction: personal agents and commerce

A personalized offer is an offer made for one customer, based on what their agent shares, with consent, and on what the merchant wants to achieve. It can be a discount, a bundle, a tailored plan or a gift. A haggl.ai sales agent makes it within the merchant’s budget and max discount.

Example: Instead of a cash discount, your sales agent offers a first-time buyer a free grinder with the espresso machine.

Read more: Upselling to AI agents with bundles

A sales agent is the AI agent that represents a business in agentic commerce. With haggl.ai, the merchant’s sales agent talks with each shopper’s agent, asks what matters and makes a personalized offer within the merchant’s budget and max discount. The merchant can see every conversation.

Example: The shopper’s agent asks whether the grinder suits espresso. Your sales agent answers and offers free shipping on the first order.

haggl.ai sales agents are live for e-commerce on Shopify Plus, with haggl.ai-assisted setup, and on Wix, through the haggl.ai app for Wix. Pilots for SaaS and utilities are scoped with us.

Read more: haggl.ai for e-commerce · Our agent shopping benchmark

A shopper’s agent is the personal AI agent that shops for a consumer: it searches, compares and asks merchants for offers, then brings its user a choice to approve. The possessive matters: the agent works for the shopper. Salesforce’s Shopper Agent is a different product, an assistant a brand runs on its own storefront.

Example: A shopper asks their assistant for a rain jacket. Acting as the shopper’s agent, it asks your sales agent what it can offer.

Read more: How to sell to AI shopping agents

Universal Commerce Protocol (UCP) is an open standard that gives platforms, AI agents and businesses a common language for commerce: catalog search, carts, checkout, identity linking and order management. Google and Shopify are among its co-developers. It accepts extensions and is open source under the Apache 2.0 license.

Read more: How the commerce protocols compare

WebMCP is a draft browser API that lets a web page give AI agents tools to call, such as searching a catalog or requesting an offer, instead of clicking through the page. The W3C Web Machine Learning Community Group publishes the draft, which is not a W3C standard. Chrome 149 added an origin trial in June 2026.

Example: The haggl.ai embed registers its offer tools through WebMCP where the browser supports it, and points other agents to REST and MCP.

Read more: The haggl.ai protocol