Support
The page offers demo requests and a "Connect with an Expert" contact route. No documentation or help centre is described.
Learn moreSize recommendations that AI shopping agents can call mid-checkout

Bold Metrics is a shopping tool. Agentic Sizing Protocol from Bold Metrics lets AI shopping agents request personalized apparel size recommendations with plain-language fit context, built on a digital-twin model and years of body data. Aimed at retailers and agent builders.
Agentic Sizing Protocol (ASP) is a product from Bold Metrics, a San Francisco company that sells body-data and fit technology to apparel brands. ASP packages that technology so an AI shopping agent can request a size recommendation on a shopper's behalf, receive the answer along with fit context, and carry on with the purchase. The problem it targets is a familiar one in agentic commerce: agents can browse and check out, yet they cannot reliably tell a shopper which size to pick. In practice, a shopper answers a handful of simple questions, and the system builds a digital twin from those answers. No photos, body scans or measuring tape are involved. The agent then calls the protocol and gets back a recommendation that goes beyond a letter size. Each measurement carries plain-language fit wording such as "just right" or "slightly snug," garments are scored to express confidence, and outlier detection supplies fallback guidance when an input looks unusual. Fit preferences can be passed along too, for example wanting a jacket looser in the chest to layer over a sweater. Responses are tuned for low token usage and quick turnaround, which matters inside an LLM loop. The protocol is described as agent agnostic, meant to work across any agent, platform or surface so that one integration serves several agentic environments. On security, retailer API credentials are kept away from the agent and from any LLM, and they never show up in responses. Bold Metrics cites more than 250 million digital twins, 12 billion body data points and 750 million fit simulations behind the model, and the page carries an endorsement from Gap Inc.'s chief technology officer. The audience is apparel retailers and brands, plus teams building shopping agents who need sizing as a callable capability. Among alternatives, it sits apart from on-site size charts and quiz widgets by being designed for machine callers. Access appears to be sales-led, through a demo request rather than self-serve signup.
The page offers demo requests and a "Connect with an Expert" contact route. No documentation or help centre is described.
Learn moreDesigned to work with any AI agent, platform or surface through a single protocol. Specific named integrations are not listed on the page.
Digital twin built from simple questions, no photos or measuring. Personalized size recommendations with per-measurement fit language, garment scoring, outlier detection with fallback guidance, fit preference support, low-token responses, agent-agnostic design and credential isolation.





