Repository navigation
Proposal: A machine-verifiable provider credential object for Local Services catalogs
#1116
mortgages11-sketch
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Posting from Neighborhood Registry (neighborhoodregistry.com). We have filed a founding-cohort nomination for the Local Services TC, and — as stated in that nomination — our first protocol contribution is the proposal below. We offer it to the TC and the community whether or not we are seated.
The gap
The Local Services charter (Issue #922) gives the TC ownership of the Service Catalog schema. The Phase 1 catalog models a service's duration, pricing, policy links, and resource requirements. It models nothing about whether the provider is legally credentialed to perform the service.
For a large, regulated slice of local services — electrical, plumbing, HVAC, general contracting, real estate, mortgage lending, law — licensure is not marketing metadata. An agent that books an unlicensed electrician creates consumer harm and platform liability that a rating does not cure. Platforms today solve this privately and incompatibly (e.g., Local Services Ads runs its own license checks against state databases behind a closed process). "Licensed & insured" as free text in a description field is unverifiable by construction. There is no open, dereferenceable way for a catalog entry to carry its credential claim and for any agent to check it.
The proposal
Add an optional
credentialobject to the provider attached to a service-catalog entry. It states a claim, cites the authority the claim comes from, says when it was verified, and gives the agent a URL where the claim resolves to a signed attestation.{ "provider": { "id": "nr:pro:example", "name": "Example Plumbing LLC", "credential": { "type": "professional_license", "license_type": "Master Plumber", "license_number": "MP000000", "issuing_authority": { "name": "Arkansas Department of Health", "jurisdiction": "US-AR", "authority_url": "https://www.healthy.arkansas.gov/" }, "status": "Active", "status_as_of": "2026-11-19", "verified_at": "2026-11-19T14:03:22Z", "validity": { "issued": "2019-05-01", "expires": "2027-04-30" }, "verification": { "method": "registry_attestation", "verifier": "Neighborhood Registry", "attestation_url": "https://neighborhoodregistry.com/verification/{id}", "attestation_format": "W3C Verifiable Credential 2.0 (Data Integrity proof)" } } } }Field discipline — the parts that make it trustworthy:
statusis the issuing authority's verbatim status string ("Active", "Active Probation", "Expired"), never an interpretation. The object reports record status only; it makes no claim about conduct, quality, or whether work was performed.status_as_of+verified_atseparate when the authority's record said this from when the verifier checked. A companion attestation endpoint serves the full dated history, so a booking flow can ask the only question that matters at dispute time: was the credential in force on the booking/service date?attestation_urlresolves to a signed, dated attestation (W3C Verifiable Credentials 2.0 data model; Data Integrity proof). The pattern deliberately mirrors AP2's signed mandates: the claim travels with its proof instead of asking the agent to trust the catalog copy.verifieris a field, not a hardcode: a state board, a platform's own screening program, or an independent registry can all issue attestations in this shape.Where it plugs in: catalog discovery can filter/rank on credential presence, jurisdiction match, and
verified_atfreshness; booking confirmation can re-resolveattestation_urlso credential state is captured as of the booking date; the same signed-attestation pattern composes with AP2 deposit mandates; and the fields map 1:1 onto schema.orghasCredentialon the provider's public web profile — one fact, two surfaces, no drift.Reference implementation and test offer
This is not a whitepaper model. We operate a production, bitemporal verification ledger — 1,113,967 dated verification records across 1.1M+ licensed professionals in all 50 states + DC (figures as of Nov 19, 2026), each citing its source authority — and 27,688,856 building permits across 57 jurisdictions, each joined to the pulling contractor's license status on the permit date. Live surfaces: a public verification API (https://neighborhoodregistry.com/verification-api/), a public ledger (https://neighborhoodregistry.com/data/verification-ledger/), an MCP server (https://mcp.neighborhoodregistry.com/mcp), and open datasets with DOIs (e.g., 10.5281/zenodo.24015303).
We offer our production attestation endpoint as a public test fixture for the TC, and we will contribute conformance test vectors — including the plausible-wrong-answer cases: an expired-but-renewed license; an "Active Probation" status vs. a bare "Probation" status (only the former asserts in-force standing); and a license status change mid-booking.
Questions for the TC
credentialbelong on the catalog provider object (our recommendation) or as a standalone discoverable resource?type,license_type,license_number,issuing_authority,status,verified_at,verification.attestation_url.verified_atage for booking-time reliance, and what re-check behavior at confirmation?Full proposal document available on request; happy to move this into a draft extension spec in the repo's format if the TC prefers.
— Neighborhood Registry (Conan Watters, Founder)
All reactions