The machine rail
When the buyer is an agent, the record answers.
AI agents already research products, compare claims, and shortlist suppliers. Listrar is the first DPP registrar designed to be read by them: a machine-readable feed and an MCP server expose the same canonical record as the passport page — no scraping, no PDFs.
The same record, machine-shaped
The feed serves the canonical record itself as structured JSON — not an export, not a copy. What an agent reads is exactly what a person sees on the passport page, at the same disclosure level.
MCP, not screen-scraping
An MCP server lets AI assistants query the register directly: look up a product, read its attributes, follow its evidence. Answers arrive with provenance attached, so an agent can tell a substantiated claim from an assertion.
No divergence, by construction
There is no second copy to maintain, so there is none to drift. When the record changes, the page and the feed change together — the register is the single source both rails resolve.
Legible where decisions are made
As agents mediate more purchasing and compliance checks, products that answer in structure get considered; products that answer in PDFs get skipped. Registering once keeps you legible on both sides of that shift.