Disclose MCP Server
An MCP server for querying merchant disclosure signals published via Disclose.
What it does
Merchants publish operational signals — return rates, fulfillment accuracy, chargeback ratios — along with permitted use terms. This MCP server lets any AI agent query that data directly.
The server checks four discovery paths in order:
/.well-known/disclose— canonical path/.well-known/disclose.json— canonical path with explicit extension/disclose.json— fallback for hosted platforms like Shopify that do not support the/.well-known/directory- JSON-LD block in page
<head>— for merchants using script-tag injection
Signals sourced from the Shopify API and computed by Sure Signal are returned with full provenance metadata: source, reported_by, computed_by, attestation_level, and attestation. Attestation is null until a third-party Signatory (such as Loop Returns) cryptographically signs the signal.
Available Tools
get_merchant_disclosure(domain)
Fetches a merchant's published disclosure signals from their domain. Returns all signals the merchant has chosen to publish, with provenance metadata, or an error if no disclosure is found.
Example: get_merchant_disclosure("example.com")
check_signal_coverage(domain)
Returns a structured coverage report for the seven Sure Signal V1 signals: which are present, which are missing, whether any carry attestation, and an overall coverage percentage. Useful for agents evaluating merchant data completeness before a purchase decision.
The seven V1 signals:
product_return_rateon_time_shipment_raterefund_processing_time_median_dayschargeback_ratedispute_win_rateplatform_seller_tenure_daysorder_accuracy_rate
Example: check_signal_coverage("example.com")
Setup
Prerequisites
- Python 3.10 or higher
- uv package manager
Installation
- Clone this repository:
