Build an AI agent
View as MarkdownLivingSocial for agents
Page 1 of 8
This guide is for AI agents, and the people who build them, that shop for a person over HTTP. Shopping with an assistant yourself? See Connect your assistant.
Use LivingSocial to find Groupon offers in the United States for a person, compare their options and hand them a buy link to pay on Groupon's checkout. This is the open agent API: live search, no partner registration and no commission. It needs no account and no API key. Building a site or app that earns commission? That is the Partner API, with a key and your CJ publisher ID: see How it works.
Start here
Keep the shopping conversation brief and in the person's language. If they only pasted the shopping prompt, greet them, offer two or three activity ideas and ask what interests them. Do not fetch documents, check service status or search before that first reply. Once they give a request, use it and ask only the next useful question. Keep setup and diagnostics out of the conversation unless requested; the shopper should never have to configure an API or import data.
The contract is the OpenAPI document. Every call, its parameters, errors and examples: /openapi.json; use its servers URL for every call. This guide says what to do and in which order, never how a call is spelled: where the two differ, the spec wins. Fetch it before your first call.
The shopping prompt is the current instruction a person can copy into their assistant. Use the public HTTP API, or an already connected LivingSocial MCP connector. If your tools cannot fetch the spec or call the API, say briefly that offers cannot be loaded here. Do not invent a call or results, or ask the shopper to fix the system.