# An agent for your catalog, not a tool to configure. *Platform · Agents* > Agents is the CommerceClarity platform surface where a catalog agent is given a target rather than a configuration. It reads the context and attribute model already agreed with the customer’s team, works out what it reads, what it fills and what it does with what it cannot source, then goes live on one category before taking the next. Agents come from a library of templates adapted to the customer, or are co-designed when no template covers the workflow. Every step runs against your attribute model and the bar your team agreed. The agent goes live on one category, then takes on the next, and a person approves anything that reaches a channel. [Book a demo](/book-a-demo) ## Blocks ### You give it the target. It builds the plan. plan *It starts from your KPIs* Bring the number you are measured on and the date you need it by. The agent works out what to read, what to fill, and what to hold for your team. - Your context, not a blank slate · It reads the product and company data you already have. · [See what it knows](/platform/context) - Your governance, from the first run · The bar and the rules it is scored against are yours. · [See how it is checked](/platform/governance) ### It grows without a new project grows *After it goes live* The agent goes live on one category with three steps, in weeks. Every month after that the Retail Engineer who put it live adds something else, and the bar it clears goes up. - Read it down its own history · Three steps in January, six now, and the date each one arrived. - Someone owns it after launch · A Retail Engineer stays on the account after go-live. ### A person approves everything that publishes human *Human in the loop* The agent scores its own work against your bar. What clears it goes to the channel, and the rest waits in a queue with the reason it was held. - The bar is yours to move · Raise it and more comes to you. Lower it and less does. - 50, not 1,200 · The queue holds only what the agent could not settle. ## Put an agent on one of your categories Bring a category, the attributes you never got to fill, and the rules it has to respect. You will see the run and what it holds back for you. [Book a demo](/book-a-demo) ## Questions ### What is a catalog agent, in practice? One workflow a retail team does by hand today, now run by software: sourcing an attribute, matching a supplier record, translating a description. Each agent has its own output, its own bar, and its own numbers. ### Do we configure the agent ourselves? No. You bring the number you are measured on and the date you need it by, and the agent works out the steps against your attribute model. ### What if the agent we need does not exist yet? Most already exist as library templates built for other customers, and adapting one to your catalog is part of the service. A brand-new agent is designed with your team and scoped separately. ### Does it run on the whole catalog at once? No. It goes live on one category, proves the number moved, and then takes on the next. Every category it finishes makes the following one cheaper, because the records are already there. ### Where does a person come into it? At the end, where it matters. The agent marks what it could not settle, a reviewer resolves those values in Grid View, and the correction becomes what the agent carries next. ## Sources - [AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) · NIST - [Regulation (EU) 2024/1689, the Artificial Intelligence Act](https://eur-lex.europa.eu/eli/reg/2024/1689/oj) · EUR-Lex - [Product data specification](https://support.google.com/merchants/answer/7052112) · Google Merchant Center - [GTIN, the GS1 global trade item number](https://www.gs1.org/standards/id-keys/gtin) · GS1 --- Canonical: https://commerceclarity.com/platform/agents Every page of this site is available as markdown: append `.md` to its path. Index: /llms.txt