Agentic Commerce Glossary

Shorthand for Agent-to-Agent — the class of protocols and interactions that enable autonomous AI agents to discover, communicate with, and transact with other agents. Google's A2A protocol and OpenAI's ACP are the leading implementations.

A machine-readable declaration published by an A2A-compatible agent that describes its capabilities, supported interactions, authentication requirements, and service endpoints — allowing other agents to discover and interact with it.

A discrete interaction between an AI agent and a merchant's commerce system. Tracking agent sessions separately from human sessions is critical for attribution and optimization in agentic commerce.

Interactions where a user's personal agent negotiates directly with a brand's sales agent to find the best product or deal, without human intervention in the loop.

A formal communication standard that allows AI agents to interoperate with one another — exchanging capabilities, negotiating transactions, and delegating tasks — without requiring human intervention. Enables brand agents to respond to user agents at scale.

A standard co-developed by OpenAI and Stripe that enables AI agents (like ChatGPT) to securely process payments and conduct transactions on behalf of users, using tokenized credentials.

A composite metric (0-100) that evaluates a brand's preparedness for AI agent commerce across 4 dimensions: AI discoverability, protocol readiness, data quality, transaction enablement, trust & verification, post-purchase experience, and measurement.

A collaborative digital environment where multiple specialized AI agents (researcher, negotiator, purchaser) work together to solve complex user goals autonomously.

The degree to which a brand's products, services, and information can be found, parsed, and understood by AI agents. Encompasses structured data quality, feed completeness, Knowledge Graph presence, and content optimization.

A brand's representation in knowledge graphs and AI training data. A strong brand entity ensures AI agents can identify, describe, and recommend the brand accurately and consistently.

A prompting technique that encourages AI to break down complex reasoning steps. In commerce, this allows agents to 'think through' a purchase decision: identifying needs, comparing options, and selecting the best value.

A secure, ephemeral state in UCP where an agent builds a cart, calculates taxes/shipping, and prepares an order for payment, without needing a full browser session.

A standardized file (usually hosted at `/.well-known/ucp`) that tells AI agents what services a merchant offers, their pricing, capabilities, and API endpoints.

A capability where an LLM can intelligently choose to call external code functions or APIs (like 'addToCart' or 'checkInventory') to compute data or take action, effectively giving the AI 'hands'.

The active, ongoing practice of improving how a brand's products and content are indexed, understood, and cited by generative AI systems. Includes structured data implementation, feed enrichment, knowledge graph presence, trust signal optimization, and content tuning for AI reasoning — distinct from one-time GEO audits.

An e-commerce architecture where the frontend (presentation layer) is decoupled from the backend (logic/data). UCP acts as a specialized 'agentic head' for headless platforms.

A structured database of entities and their relationships used by AI systems to understand the world. For e-commerce, being present in knowledge graphs means AI agents 'know' your brand exists and can reference it.

Return policies, shipping tables, warranty information, and compliance data formatted in structured, parseable formats (JSON-LD, microdata) rather than buried in PDFs or legal pages that AI agents can't read.

Salesforce Commerce Cloud's Open Commerce API and Shopper Commerce API. ForkPoint's cartridge bridges these traditional APIs to the Universal Commerce Protocol.

A secure component in the UCP architecture that processes payments. Agents hand off the final payment step to this secure handler (or generate a payment link) to ensure PCI compliance.

A structured data file containing product information (titles, descriptions, prices, images, inventory) syndicated to platforms like Google Merchant Center. Feed quality directly impacts AI agent recommendations.

The degree to which a merchant has implemented the agent commerce protocols (UCP, MCP, A2A) needed for AI agents to transact with their store. A key dimension of agentic commerce readiness.

A technique where an AI retrieves relevant data from an external source (like a UCP product catalog) to ground its answers, preventing hallucinations and ensuring up-to-date pricing/availability.

Machine-readable markup (typically JSON-LD using Schema.org vocabulary) embedded in web pages that helps AI agents understand product attributes, pricing, availability, reviews, and organizational information.

Indicators that help AI agents assess a brand's reliability: aggregated reviews, merchant verification, return policy clarity, shipping transparency, and compliance certifications. Agents prioritize brands with strong trust signals.

The ability of an AI model to perform a task (like 'find a gift for a 5-year-old') without seeing specific examples beforehand. High-quality structured data (like UCP manifests) improves zero-shot performance.