Anthropic Releases Blueprints for Building Commerce Agents

These blueprints include the frameworks, patterns, and guardrails necessary for engineering teams to get agents operational within a matter of days.

The provided implementation examples include reference implementations for shopping agents and merchant agents tailored for retail, travel, telecommunications, and ticketing platforms.

Shopping agents support users in searching for, comparing, and purchasing products based on natural language requests.

They are equipped with guardrails to restrict pricing and product information to actual catalog data, a design intended to avoid inappropriate upselling.

Conversely, merchant agents are designed to assist store operators.

These agents generate responses based on store-side data, such as suggesting items for discounts to clear inventory.

Changes proposed by the agent are structured to be applied only after final human approval.

This solution is available in environments such as the Claude API, Amazon Bedrock, Microsoft Foundry, and Google Cloud Vertex AI.

Additionally, partner companies including Accenture, Mastercard, and Visa are collaborating to make these tools available to their clients.