AMP Evidence Atlas · Public-safe beta

1,374papers indexed

9,263peptide summaries

108,761activity rows aggregated

128,976audit rows aggregated

28,813conflict statuses

Normalized database

Six layers, one rights-safe read model.

System provenance, source governance, discovery catalogs, evidence aggregates, evaluation items and stable API views are separated so public clients cannot accidentally cross into restricted raw records.

system

Release identity, provenance checksum and scope flags.

system_release

governance

Per-source redistribution decisions and terms links.

governance_source_rights

catalog

Derived paper and peptide discovery indexes.

catalog_papercatalog_peptidecatalog_peptide_sequencecatalog_peptide_endpointcatalog_peptide_target_examplecatalog_peptide_evidence_tiercatalog_peptide_paper

evidence

Aggregate audit outcomes; no row-level source comparison.

evidence_audit_aggregateevidence_difference_category

evaluation

Project-authored benchmark questions and answer keys.

evaluation_benchmark_item

api

Stable read views used by API and AI/MCP clients.

api_peptide_summaryapi_paper_summary

Catalog

Browse peptides and papers.

Search and filter derived summaries. Counts describe the internal evidence corpus, while the public response intentionally omits copied source fields and row-level comparisons.

Loading public-safe summaries…

Database audit

See disagreement patterns, not verdicts.

The atlas preserves source-verified, source-conflict and other review states. Here they are exposed only as aggregates and must not be read as automatically human-confirmed database errors.

Audit status by database

Loading…

Most frequent difference categories

Interpretation boundary. Multi-agent agreement is not a human gold standard. Stratified manual validation and source-permission follow-up remain release blockers for the unrestricted resource.

Evaluation

Grounding benchmark preview.

Forty project-authored questions test whether a model can recover activity values, sequence details, database–paper conflicts and evidence limitations. This pilot is not yet a resource-grade gold standard.

One corpus, two stable interfaces.

Use hierarchical REST endpoints for applications, or let an AI client discover and invoke ten bounded read-only tools through the same-origin Streamable HTTP MCP endpoint.

REST API v1

Canonical endpoints return a consistent data / meta / links envelope. Legacy flat paths remain aliases.

GET/api/v1/system/stats

GET/api/v1/catalog/peptides

GET/api/v1/catalog/papers

GET/api/v1/evidence/audit-summary

GET/api/v1/evaluation/benchmark

GET/api/v1/schema/database

Open OpenAPI 3.1 schema →

Streamable HTTP MCP

Stateless endpoint at /api/mcp, supporting modern 2026-07-28 discovery and legacy 2025-11-25 initialization. No authentication or write tools.

{"mcpServers":{"amp-evidence-atlas":{"url":"https://amp-evidence-atlas.daoyu7974.chatgpt.site/api/mcp"}}}

Server descriptor

curl -X POST https://amp-evidence-atlas.daoyu7974.chatgpt.site/api/mcp \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -H 'MCP-Protocol-Version: 2026-07-28' \ -H 'Mcp-Method: tools/list' \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

atlas.describe

Describe AMP Evidence Atlas

Return release scope, aggregate statistics and the normalized public database hierarchy.

atlas.search

Search the public-safe catalog

Search peptide names, displayed sequence variants, paper IDs and DOIs.

atlas.list_peptides

List peptide summaries

Browse and filter rights-safe derived peptide summaries with bounded pagination.

atlas.get_peptide

Get one peptide

Return the exact public-safe summary for a peptide name.

atlas.list_papers

List paper summaries

Browse paper identifiers and project-created review/count metadata.

atlas.get_paper

Get one paper

Return the exact public-safe summary for a paper ID.

atlas.get_audit_summary

Get aggregate audit outcomes

Return database/status and difference-category aggregates, optionally filtered.

atlas.get_benchmark

Get benchmark items

Return bounded project-authored grounding benchmark items and answer keys.

atlas.get_rights_policy

Get rights policy

Return source-rights decisions, excluded components and open permission follow-up.

atlas.get_database_schema

Get public database schema

Return layers, entities, views and relationships of the normalized rights-safe read model.

Live API console

Run a rights-safe API request.

Live MCP tool call

Call a modern stateless MCP tool.

Rights & release governance

Public by derivation, limited by design.

This beta publishes project-created metadata and aggregates while the immutable full evidence package stays internal until permissions, validation and publication review close.

Source	Assessment	Public decision	Reference
APD6	no_explicit_broad_redistribution_license_found	raw_APD6_export_not_cleared_for_public_rehosting	Terms ↗
CAMP	open_access_is_not_a_redistribution_license	raw_CAMP_export_not_cleared_for_public_rehosting	Terms ↗
DBAASP	internally_contradictory_distribution_terms	raw_DBAASP_export_not_cleared_for_public_rehosting	Terms ↗
DRAMP	conditionally_cleared_under_CC_BY_4_0	DRAMP_general_and_clinical_data_may_be_reused_with_attribution_but_patent_AMPs_are_excluded	Terms ↗
dbAMP	academic_use_permission_without_express_public_redistribution_right	raw_dbAMP_export_not_cleared_for_public_rehosting	Terms ↗

Never exposed here

bulk source-database record mirrors
source-database record names, values, units and source IDs
primary article full text, PDFs, tables and figures
bulk row-level database audit comparisons
DRAMP patent AMP content

Open permission follow-up

APD6CAMPDBAASPdbAMPDRAMP patent AMPs

Machine-readable policy →