GenesisPay index agent

Last updated: 15 November 2026

The GenesisPay index agent records which online stores publish a UCP profile, so people and AI assistants can find stores that support agent shopping. It is run by Genesis Technologies FlexCo (legal disclosure).

Optional merchant offer research

An optional research pilot reads a store’s homepage and up to two linked public offer pages on that store’s own host, about once per day with scheduling jitter. It keeps explicitly published codes, a short text excerpt, the source address, a text hash and observation times for review and optional display in catalogue answers. Codes and excerpts can contain publicly stated names, such as influencer codes. Full pages are discarded and no cookies are sent. Each displayed observation keeps its source and time; checkout acceptance is unconfirmed. It never affects ranking.

Requests start at least one second apart. Each request can follow at most two redirects, with robots.txt checked for every target. Disallow the relevant paths or all paths for GenesisPayIndex, return 401/403, or contact us to opt out. Expired evidence is removed during bounded daily maintenance; cleanup resumes if its maintenance budget is exhausted. Rows leave the dataset after 30 days outside its scope. Collection and catalogue display are controlled separately. Its request identifier is:

GenesisPayIndex/0.6 (+https://genesispay.finance/ucp/index-agent; reads public UCP profiles, store details, policy addresses and public offer pages, never checkout)

What it does

  • Per store, one GET /robots.txt and, if robots.txt allows it, one GET /.well-known/ucp. The product catalog is read as described below.
  • When its store-details lane is enabled, a store with a Shopify UCP endpoint is also read at /meta.json. Only country and ships_to_countries are kept; city, province, description and identifiers are discarded. It runs every 14 days for prioritised stores, 30 for the long tail. Opt out with Disallow: /meta.json for GenesisPayIndex, answer 401/403, or ask us to remove the store. If source comparison requires it, one tokenless, read-only query for localization.availableCountriesalso reads /api/2025-07/graphql.json, only when robots.txt allows it. It opens no cart. The collection lane runs only after its approval gates are met.
  • When policy-page collection is enabled, the index checks four policy-page addresses using HEAD; redirects and retries can add requests. The addresses are /policies/legal-notice, /policies/refund-policy, /policies/shipping-policy and /policies/contact-information. Each path must be allowed by robots.txt. Only addresses are kept, never page content. Requests start at least one second apart, including retries and redirects.
  • When domain-date collection is enabled, the index asks the domain registry's public RDAP service, selected only from IANA's bootstrap file (data.iana.org/rdap/dns.json, refreshed at most once a day). Only the registration date is kept; entities, contacts and other answer fields are discarded. Successful lookups repeat after 180 days (with scheduling jitter); failures retry sooner. Requests to one registry start at least two seconds apart. Stores under a suffix the public suffix list marks as shared (such as myshopify.com or uk.com) are not looked up. robots.txt does not apply to registries; ask us to remove the store to stop its lookup.
  • Profile probes are scheduled at 7–90 days depending on priority and previous results, with ±10% scheduling jitter; failed attempts can retry sooner. Backlogs, failures or maintenance interruptions can delay checks.
  • If the profile advertises a UCP agent endpoint, the agent reads the store's product catalog through that endpoint the way the UCP specification describes: it calls the catalog tools search_catalog and, where needed, lookup_catalog, and names its own profile on every call so the store can decide what to answer. It reads a page of up to 100 products at a time, with a pause between pages, and reads the whole catalog again about once a day (less often for small stores).
  • The catalogue reader never logs in, never opens a cart and never starts a checkout. It calls no cart, checkout, order or payment tool, and its profile declares none.
  • A one-off, bounded pilot uses GenesisPayIndex to read at most 60 product pages and 30 homepages, and the UCP catalogue with AT, DE and US country contexts, to measure store-published shipping costs and delivery times. It reads only embedded structured facts, not policy text, respects robots.txt and blocks, and creates no cart; this pilot does not display its results in the catalogue.
  • A separate cart probe (GenesisPayCartProbe) briefly creates a test cart to check the price, discount codes, shipping costs and delivery time.
  • The profile check leaves a store that disallows it, or answers 401, 403 or 429, alone for 30 days. A store-details 429 or temporary failure retries after 1, 3, then 14 days, with Retry-After able to extend the delay up to 14 days. The agent does not change addresses or user agents to get around a block.

User-Agent

GenesisPayIndex/0.5 (+https://genesispay.finance/ucp/index-agent; reads public UCP profiles, store details and the addresses of store policy pages, never checkout)
GenesisPayIndex/0.6 (+https://genesispay.finance/ucp/index-agent; reads public UCP profiles, store details, policy addresses and public offer pages, never checkout)
GenesisPayIndex GenesisPayCartProbe (+https://genesispay.finance/ucp/index-agent; test carts only, never orders)

The robots.txt tokens are GenesisPayIndex and GenesisPayCartProbe.

What is stored

The public facts of the profile: its protocol version, the endpoint it advertises, the capabilities and payment handlers it declares, and a hash of the file. Plus the result of the last check and when it ran. From the catalog: each product's id, title, description as plain text, page URL, listed price, first picture URL, categories, tags, number of variants and whether it is available, as the store published them. No page content beyond that is kept, except the bounded shipping-pilot evidence described below. Public business facts may be personal data when a store is run by an individual. The products are shown to people and AI assistants through GenesisPay's read-only catalog, which links back to the store; a purchase happens on the store's own site.

Shipping pilot

For this pilot we process structured shipping facts, conditions, source addresses and read times published by stores to assess shipping-information quality. Full pages and responses are processed only in memory and then discarded. Only the operator and reviewer (Tom and Claude/Codex sessions running QA on the operator host) receive bounded evidence (at most 1 KiB per fact, 8 KiB per store) and measurements, including the store-linked request ledger, outside the source repository; store-linked records are deleted within seven days of collection, including after an interrupted run. We collect no customer data, use no cookies or customer sessions, and store no session identifiers. Only reviewed anonymous aggregates that do not single out a store (no one-store strata) may enter the report. Stores can object through robots.txt for GenesisPayIndex, HTTP 401/403 or the contact on this page; we stop collection and delete their evidence.

Country, policy-address and domain retention

Country and shipping-country declarations and policy-page addresses are deleted during daily maintenance after 30 days outside our Shopify-profile collection scope, unless the store returns. Domain mappings are deleted after 30 days outside the index; domain-registration records are removed when no retained store mapping refers to them. Declarations and policy-page addresses are also cleared after 60 days without a fresh observation. The departure periods start when maintenance first records the changed status. Re-check intervals depend on priority and previous results: profile probes are scheduled at 7–90 days, with ±10% scheduling jitter. Backlogs, failures or maintenance interruptions can delay checks and deletion.

Profiles, products and endpoint-health history follow separate rules. Profile observations have no fixed automatic deletion period. UCP products missing from a completed catalogue sync are marked removed rather than deleted; feed products follow source expiry and withdrawal rules. Raw endpoint-health samples are kept for 30 days by default (configurable, never less); daily rollups and current health rows have no fixed automatic deletion period. The declaration and address periods above do not apply to these records.

How to opt out

Add this to your robots.txt to stop the profile check, the catalogue read and the cart probe:

User-agent: GenesisPayIndex
Disallow: /.well-known/ucp

User-agent: GenesisPayCartProbe
Disallow: /.well-known/ucp

To keep the profile check but stop the catalog read, disallow the path of your UCP endpoint instead (on the host that serves it). Or email [email protected] with your store's domain, and we remove it and its products from the index.

Disallow: /policies/ stops only the policy-page requests.Disallow: / stops all requests to the store.

Agent profile

The agent's UCP platform profile is at /ucp/index-agent/profile.json. It is the identity of this reader only: it declares the catalog read capabilities (dev.ucp.shopping.catalog.search, dev.ucp.shopping.catalog.lookup and the Shopify catalog extensions) and nothing that could buy. The agent names that URL with a short ?v= query derived from the profile's content, because some platforms cache a fetched profile longer than its max-age; a changed profile therefore gets a fresh URL, and the query changes nothing in the answer. An agent that buys on someone's behalf will use a separate profile.