OP here. I'm exploring a new web standard called procurement.txt (robots.txt for agent commerce). It supports everything from simple setups ("email @store.example”) to sophisticated automation standards (UCP, ACP). The pattern would be to check /procurement.txt on any commercial domain and then act accordingly.
I think the internet will adjust to accommodate agents, just like it did for mobile. Agents are the fastest growing website user segment today (like mobile used to be). Others are trying similar things for content (llms.txt).
This should work for any transaction type, but I had long-tail B2B procurement in mind since it's the least supported today by current attempts. Agents will handle discovery and evaluation in the future (at a minimum).
I have found scraping the web to be slow and expensive, and many B2B websites are unusable. Merchants should be properly incentivized if agents adopt the standard.
Early version of the spec is live. I'm seeing ~90% reduction in token usage, but there are likely gaps I've overlooked. Open to collaborators and feedback.
OP here. I'm exploring a new web standard called procurement.txt (robots.txt for agent commerce). It supports everything from simple setups ("email @store.example”) to sophisticated automation standards (UCP, ACP). The pattern would be to check /procurement.txt on any commercial domain and then act accordingly.
I think the internet will adjust to accommodate agents, just like it did for mobile. Agents are the fastest growing website user segment today (like mobile used to be). Others are trying similar things for content (llms.txt).
This should work for any transaction type, but I had long-tail B2B procurement in mind since it's the least supported today by current attempts. Agents will handle discovery and evaluation in the future (at a minimum).
I have found scraping the web to be slow and expensive, and many B2B websites are unusable. Merchants should be properly incentivized if agents adopt the standard.
Early version of the spec is live. I'm seeing ~90% reduction in token usage, but there are likely gaps I've overlooked. Open to collaborators and feedback.