compare_products
ChatGPTPer-analyte evidence matrix for up to 8 analytes across up to 5 public products, using latest or explicitly named lots. Missing analytes have status not_tested; unavailable products have status record_unavailable. Optional limit and unit produce a strictly-below threshold check. Non-detects and incompatible units have an indeterminate threshold verdict.
fetch
ChatGPTFull text of one public LabTested product record, including applicable recalls, published lots, reported results and its canonical source URL. Private, unshared or undiscoverable records return an unavailable response.
get_lot_results
ChatGPTComplete published results for one public product lot: reported values, units, printed limits, laboratory verdicts, result types, statuses, laboratory and accreditation text, test date, provenance and source URLs. An unknown supplied lot returns lot_not_found with no substitute lot. Omitted lot selects the most recent published lot. Result states distinguish not_detected, below_reporting_limit, measured_within_limit, measured_above_limit, reported_without_limit, reported_without_verdict, not_tested and record_unavailable. Within/above limit reflects the laboratory verdict. A reporting bound is not an exact measurement or proof of absence. Missing testing has status not_tested, not a failed result.
get_original_coa
ChatGPTPublic original certificate-of-analysis links and available SHA-256 fingerprints for one product lot. Only documents the brand shared publicly are included. The response contains source links and provenance rather than document file bytes.
get_product_testing
ChatGPTPublished testing overview for one public product: lot numbers, dates, laboratory attribution, covered result families, status counts and applicable recalls. Result states distinguish not_detected, below_reporting_limit, measured_within_limit, measured_above_limit, reported_without_limit, reported_without_verdict, not_tested and record_unavailable. Within/above limit reflects the laboratory verdict. A reporting bound is not an exact measurement or proof of absence. Missing testing has status not_tested, not a failed result.
render_evidence_card
ChatGPTDefault interactive display for requested test results, lab results and result comparisons. Product view displays the full published results for one public product and its latest or explicitly named lot. Compare view displays a product/analyte matrix with optional lots and threshold filters. Search view displays up to 10 public product matches. The card reloads consent-gated records from identifiers; arbitrary result values are not accepted. Unknown lots remain unavailable without another lot being substituted.
search
ChatGPTCitable public LabTested product records matching a brand, product or ingredient query. Each result contains a public record identifier, product title and canonical source URL. The catalog is restricted to approved, opted-in public evidence.
search_products
ChatGPTSearches the approved, opted-in public LabTested catalog by brand, product name, ingredient or category. Optional filters cover result families and analyte thresholds on the latest lot. Results contain public product identifiers, lot dates, tested families and source URLs. Pagination examines at most 20 candidates per page and returns next_cursor. Catalog absence is not evidence that a product was untested.
compare_products
ClaudePer-analyte evidence matrix for up to 8 analytes across up to 5 public products, using latest or explicitly named lots. Missing analytes have status not_tested; unavailable products have status record_unavailable. Optional limit and unit produce a strictly-below threshold check. Non-detects and incompatible units have an indeterminate threshold verdict.
fetch
ClaudeFull text of one public LabTested product record, including applicable recalls, published lots, reported results and its canonical source URL. Private, unshared or undiscoverable records return an unavailable response.
get_lot_results
ClaudeComplete published results for one public product lot: reported values, units, printed limits, laboratory verdicts, result types, statuses, laboratory and accreditation text, test date, provenance and source URLs. An unknown supplied lot returns lot_not_found with no substitute lot. Omitted lot selects the most recent published lot. Result states distinguish not_detected, below_reporting_limit, measured_within_limit, measured_above_limit, reported_without_limit, reported_without_verdict, not_tested and record_unavailable. Within/above limit reflects the laboratory verdict. A reporting bound is not an exact measurement or proof of absence. Missing testing has status not_tested, not a failed result.
get_original_coa
ClaudePublic original certificate-of-analysis links and available SHA-256 fingerprints for one product lot. Only documents the brand shared publicly are included. The response contains source links and provenance rather than document file bytes.
get_product_testing
ClaudePublished testing overview for one public product: lot numbers, dates, laboratory attribution, covered result families, status counts and applicable recalls. Result states distinguish not_detected, below_reporting_limit, measured_within_limit, measured_above_limit, reported_without_limit, reported_without_verdict, not_tested and record_unavailable. Within/above limit reflects the laboratory verdict. A reporting bound is not an exact measurement or proof of absence. Missing testing has status not_tested, not a failed result.
render_evidence_card
ClaudeReturns an interactive evidence card for product, comparison or search results, suited to requests to see full results or a comparison in an MCP Apps host. Product view displays the full published results for one public product and its latest or explicitly named lot. Compare view displays a product/analyte matrix with optional lots and threshold filters. Search view displays up to 10 public product matches. The card reloads consent-gated records from identifiers; arbitrary result values are not accepted. Unknown lots remain unavailable without another lot being substituted. The response prepares card data; display is controlled by the host.
search
ClaudeCitable public LabTested product records matching a brand, product or ingredient query. Each result contains a public record identifier, product title and canonical source URL. The catalog is restricted to approved, opted-in public evidence.
search_products
ClaudeSearches the approved, opted-in public LabTested catalog by brand, product name, ingredient or category. Optional filters cover result families and analyte thresholds on the latest lot. Results contain public product identifiers, lot dates, tested families and source URLs. Pagination examines at most 20 candidates per page and returns next_cursor. Catalog absence is not evidence that a product was untested.