We open-sourced Facet, the protocol layer for agent-ready businesses. When an AI agent visits a website, it needs three things the open web does not currently provide: a way to identify itself, a way to discover what is for sale in machine-readable form, and a way to be trusted by the merchant on the other side. Facet ships the protocol that answers all three, plus a reference implementation, an MCP-native Agent Terminal, a schema generator that maps Shopify, WooCommerce, and NetSuite catalogs into typed manifests, and a reputation knowledge graph. The Universal Business Index has 9.4M businesses indexed and is closing in on 10M; the knowledge graph layer is ingesting 24/7 and on track for parity by end of week. Eight business primitives cover the agent-action shape of every commerce surface on the web: catalog, paywalled content, subscription, booking, date-bound inventory, auction, quote/RFQ, and credentialed (overlay). A thin ninth, view + handoff, covers the ~25% of businesses (cafes, walk-in retail, civic and public goods) that don't transact through a website but need to be discoverable. The schema generator and the reputation registry are vertical-agnostic from day one. Facet was a product that came out of a parallel project, now spinning out with a clean cap table. We are fundraising raising to ship the next four protocol layers: agent WAF, signed Ed25519 response provenance v1, reputation v1, and the user-guided schema generator across all businesses on the web. The repo is at github.com/facet-llc/spec. curl -s https://facet.llc returns the 402 protocol response. That is the rail talking back. If the thesis maps, the inbox is open