Is your feature request related to a problem? Please describe.
The UCP initiative elegantly solves the N×N integration problem for cart management and secure checkout in agentic commerce.
However, there is a critical bottleneck in the Discovery Phase of the shopping funnel.
Before an AI agent can execute a UCP checkout, it must discover the product, compare prices, and check inventory across hundreds of items. If an agent relies on traditional REST APIs, MCP, or heavy JSON-LD/HTML schemas to scan a merchant's catalog of 30,000+ items, the token consumption (and associated LLM inference cost) is astronomical. UCP is brilliant for stateful, secure execution, but it is fundamentally too heavy and expensive for stateless, exploratory discovery by LLMs.
Describe the solution you'd like
I propose officially defining a bridge to a lightweight, text-based discovery layer to act as the "top of the funnel" for UCP.
I am the author of CommerceTXT - a CC0, token-optimized, flat-text discovery protocol that reduces AI agent token consumption by ~95% compared to JSON/HTML.
The Proposed Pipeline:
-
Discovery (Stateless, Fast, Cheap): The AI agent reads the merchant's commerce.txt to instantly understand inventory, pricing, and semantic logic for pennies in compute.
-
Execution (Stateful, Secure): Once the user commits to a purchase, the agent reads a UCP endpoint directive within the text file and hands off the transaction to the UCP framework.
By standardizing this hand-off, UCP handles the complex checkout, while the text protocol handles the cheap discovery.
Describe alternatives you've considered
-
Expanding UCP APIs for full catalog discovery: This mixes public, stateless discovery data with stateful transaction logic, creating unnecessary bloat and API overhead for merchants.
-
Relying on standard HTML scraping or JSON-LD/Schema.org: We benchmarked this extensively. The "Syntax Tax" (brackets, quotes, nested objects) in JSON wastes 20-30% of the LLM context window. It is too verbose for agentic shopping at scale.
Additional context
Proof of Scale: We recently benchmarked the entire 30,000+ item IKEA US catalog using CommerceTXT. We achieved a ~24% total token reduction compared to minified JSON (saving millions of tokens) while maintaining perfect hierarchical routing for AI agents.
Strategic Value: With competing standards emerging (like OpenAI's Agentic Commerce Protocol), the standard that wins global adoption will be the one that is most cost-effective for LLM infrastructure. Offering an ultra-cheap discovery layer paired with UCP's execution layer makes adoption a no-brainer for both AI providers and small merchants (who just drop a .txt file on their server).
I would love to discuss how the CommerceTXT open working group can align with UCP's roadmap.
Code of Conduct
Is your feature request related to a problem? Please describe.
The UCP initiative elegantly solves the N×N integration problem for cart management and secure checkout in agentic commerce.
However, there is a critical bottleneck in the Discovery Phase of the shopping funnel.
Before an AI agent can execute a UCP checkout, it must discover the product, compare prices, and check inventory across hundreds of items. If an agent relies on traditional REST APIs, MCP, or heavy JSON-LD/HTML schemas to scan a merchant's catalog of 30,000+ items, the token consumption (and associated LLM inference cost) is astronomical. UCP is brilliant for stateful, secure execution, but it is fundamentally too heavy and expensive for stateless, exploratory discovery by LLMs.
Describe the solution you'd like
I propose officially defining a bridge to a lightweight, text-based discovery layer to act as the "top of the funnel" for UCP.
I am the author of CommerceTXT - a CC0, token-optimized, flat-text discovery protocol that reduces AI agent token consumption by ~95% compared to JSON/HTML.
The Proposed Pipeline:
Discovery (Stateless, Fast, Cheap): The AI agent reads the merchant's commerce.txt to instantly understand inventory, pricing, and semantic logic for pennies in compute.
Execution (Stateful, Secure): Once the user commits to a purchase, the agent reads a UCP endpoint directive within the text file and hands off the transaction to the UCP framework.
By standardizing this hand-off, UCP handles the complex checkout, while the text protocol handles the cheap discovery.
Describe alternatives you've considered
Expanding UCP APIs for full catalog discovery: This mixes public, stateless discovery data with stateful transaction logic, creating unnecessary bloat and API overhead for merchants.
Relying on standard HTML scraping or JSON-LD/Schema.org: We benchmarked this extensively. The "Syntax Tax" (brackets, quotes, nested objects) in JSON wastes 20-30% of the LLM context window. It is too verbose for agentic shopping at scale.
Additional context
Proof of Scale: We recently benchmarked the entire 30,000+ item IKEA US catalog using CommerceTXT. We achieved a ~24% total token reduction compared to minified JSON (saving millions of tokens) while maintaining perfect hierarchical routing for AI agents.
Strategic Value: With competing standards emerging (like OpenAI's Agentic Commerce Protocol), the standard that wins global adoption will be the one that is most cost-effective for LLM infrastructure. Offering an ultra-cheap discovery layer paired with UCP's execution layer makes adoption a no-brainer for both AI providers and small merchants (who just drop a .txt file on their server).
I would love to discuss how the CommerceTXT open working group can align with UCP's roadmap.
Code of Conduct