Show HN: Shelf Protocol – Robots.txt for Commerce

github.com

1 point by Lui371 15 minutes ago

My project helps merchants, consumers, and ai agents. We are at the point ai agents will be buying things for consumers. My product verifies the merchant, highlights the store to agents, brings stores and their products into one registery and gives the ai agent speed by calling my registry instead of the agent going searching each merchant one at a time. I have a function "can_buy()" that does a few things. It gives agents an understanding of what they can and can't do. It helps them know if a merchant allows ai agents to buy without a human involved, it lets the agent know my product prove the store is really who it says it is and not a fake site, and it lets the agent know how much it can spend before it has to ask a human for permisssion to proceed. For the consumer it lets your agent do things autonomously without always asking for permission to buy if its under a certain amount. And the agent is going to a store that is verified by my system. For the merchant this gives you another avenue to sell. Where things are heading agents using this system will skip stores that it can't verify. With guardrails that will most likely be in place for agents if you're not listed in this system you'll be invisible. It's one DNS record, free and five minutes. It gives you control on what agents see. Such as your products, how much you allow them to purchase and you are verified as the domain owner of your store for AI agents. I didn't want to launch with an empty directory so I pre-filled the registry with already public product data from real stores on Shopify. I have 8 real merchants and 834 real products. They aren't verified and there is a 0 spending limit at this time. I need the merchants to claim their listing first. It's early and I would love some feedback, especially on the shelf.json schema. I would also like your thoughts on whether DNS-TXT is the right verification long term.