Blog
News and updates about the Open Storefront Protocol and the Cresva developer surface.
The discovery document, quoted whole. Three of its fields say false with a reason attached rather than guessing, and the one an agent most wants to be true is the one that refuses.
Three of seven components come back null rather than as a plausible middle number, and measured_weight says how much of the score rests on evidence. The live response, pasted.
Five weighted categories, every check named, and two things the earlier post got wrong: the suite reads data the platform already holds rather than calling your endpoints, and the re-run job is currently off.
The old acronym now belongs to a specification Stripe and OpenAI publish. Nothing inside the protocol was renamed, and both version headers are still sent.
MCP, ACP, AP2 and OSP, read from their own documentation. They are not four candidates for one job, and the one place two of them genuinely disagree is about who holds the money.
A deep dive into how OSP computes trust scores across independent storefronts without a central authority. Covers Bayesian aggregation, decay functions, and cross-platform reputation portability.
Traditional REST APIs were designed for human-driven UIs. Agent commerce demands structured negotiation, semantic product understanding, and machine-readable trust signals. We explain why OSP exists.
Trust is the fundamental primitive of agent commerce. This post covers the cryptographic verification, behavioral scoring, and transparency mechanisms that make OSP transactions trustworthy.
