TrialMCP Pack: MCP Servers for Physical AI Oncology Clinical Trial Systems

Fuente: Zenodo
Salvato in:
Dettagli Bibliografici
Autori principali: Kevin Kawchak, Claude, Codex
Natura: Recurso digital
Pubblicazione: Zenodo 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866901896317370368
author Kevin Kawchak
Claude
Codex
author_facet Kevin Kawchak
Claude
Codex
contents <h2>Summary</h2> <p>This release implements code fixes and structural improvements based on the @codex v0.1.1 peer-review recommendations, advancing the TrialMCP Pack toward full-length paper publication readiness. Key changes include token expiry enforcement, shared error taxonomy across all servers, input validation with SSRF prevention, health/readiness endpoints, policy decision traces, pagination guards, dataset governance artifacts, and expanded test coverage from 30 to 39 tests (all passing).</p> <h2>Features</h2> <ul> <li><strong>Shared common module</strong> (<code>servers/common/</code>): Machine-readable error codes (<code>AUTHZ_DENIED</code>, <code>VALIDATION_FAILED</code>, <code>NOT_FOUND</code>, etc.), input validation utilities (FHIR ID format, DICOM UID format, URL rejection), and health endpoint helper -- addressing peer-review recommendation 1.1 (cross-server contracts).</li> <li><strong>Token expiry enforcement</strong>: Tokens now include <code>expires_at</code> timestamps and are automatically rejected after expiration -- addressing peer-review recommendation 1.2 (token lifecycle).</li> <li><strong>Policy decision traces</strong>: <code>authz_evaluate</code> responses now include a trace of all matching rules for auditable authorization decisions -- addressing peer-review recommendation 1.2 (policy decision traces).</li> <li><strong>Input validation</strong>: FHIR resource IDs and DICOM UIDs are validated against format patterns; URLs are rejected at the validation layer with <code>VALIDATION_FAILED</code> error codes -- addressing peer-review recommendations 1.1 and 1.4 (security controls).</li> <li><strong>Health endpoints</strong>: All five MCP servers expose a <code>health</code> tool returning server name, version, status, and timestamp -- addressing peer-review recommendation 1.7 (operational robustness).</li> <li><strong>Pagination guard</strong>: FHIR search now enforces <code>max_results</code> to prevent unbounded result sets -- addressing peer-review recommendation 1.3 (query guardrails).</li> <li><strong>Dataset governance</strong>: <code>datasets/README.md</code> with data dictionary, provenance, and known limitations; <code>datasets/manifest.json</code> with SHA-256 checksums -- addressing peer-review recommendation 3.1.</li> <li><strong>Repository-level files</strong>: <code>SECURITY.md</code>, <code>CONTRIBUTING.md</code>, <code>CITATION.cff</code> for publication readiness -- addressing peer-review repository-level recommendations.</li> <li><strong>3 new Mermaid diagrams</strong> in README: peer-review development lifecycle, quality assurance cycle, and MCP PAI oncology trial lifecycle.</li> <li><strong>Expanded test suite</strong>: 9 new tests including token expiry, encoded SSRF variants, health endpoints, policy traces, and cross-server integration trace.</li> <li><strong>Contributor icons</strong>: @kevinkawchak, @claude, and @codex profile images in README Team section.</li> </ul> <h2>Contributors</h2> <p>@kevinkawchak @claude @codex</p> <h2>Notes</h2> <ul> <li>All 39 tests pass. Test suite expanded from 30 (v0.1.0) to 39.</li> <li>All five MCP server versions updated to 0.2.0.</li> <li>This release addresses peer-review recommendations from <code>peer-review/trialmcp_pack_v0.1.1_peer_review.md</code> sections 1.1 (shared contracts), 1.2 (token lifecycle, policy traces), 1.3 (pagination), 1.4 (DICOM security), 1.7 (health endpoints), 3.1 (dataset governance), and repository-level recommendations (SECURITY.md, CONTRIBUTING.md, CITATION.cff).</li> <li>Remaining peer-review recommendations (1.3 de-identification profiles, 1.5 signature support, 1.6 W3C PROV alignment, 2.1 client state machine, 4.1-4.3 adversarial tests, paper/ directory, CI/CD workflows) are documented as future work.</li> <li>Python 3.11+ required. No new external dependencies.</li> </ul>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18870616
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle TrialMCP Pack: MCP Servers for Physical AI Oncology Clinical Trial Systems
Kevin Kawchak
Claude
Codex
model-context-protocol
oncology
clinical-trials
physical-ai
robotics
fhir
dicom
audit
hipaa
mcp
<h2>Summary</h2> <p>This release implements code fixes and structural improvements based on the @codex v0.1.1 peer-review recommendations, advancing the TrialMCP Pack toward full-length paper publication readiness. Key changes include token expiry enforcement, shared error taxonomy across all servers, input validation with SSRF prevention, health/readiness endpoints, policy decision traces, pagination guards, dataset governance artifacts, and expanded test coverage from 30 to 39 tests (all passing).</p> <h2>Features</h2> <ul> <li><strong>Shared common module</strong> (<code>servers/common/</code>): Machine-readable error codes (<code>AUTHZ_DENIED</code>, <code>VALIDATION_FAILED</code>, <code>NOT_FOUND</code>, etc.), input validation utilities (FHIR ID format, DICOM UID format, URL rejection), and health endpoint helper -- addressing peer-review recommendation 1.1 (cross-server contracts).</li> <li><strong>Token expiry enforcement</strong>: Tokens now include <code>expires_at</code> timestamps and are automatically rejected after expiration -- addressing peer-review recommendation 1.2 (token lifecycle).</li> <li><strong>Policy decision traces</strong>: <code>authz_evaluate</code> responses now include a trace of all matching rules for auditable authorization decisions -- addressing peer-review recommendation 1.2 (policy decision traces).</li> <li><strong>Input validation</strong>: FHIR resource IDs and DICOM UIDs are validated against format patterns; URLs are rejected at the validation layer with <code>VALIDATION_FAILED</code> error codes -- addressing peer-review recommendations 1.1 and 1.4 (security controls).</li> <li><strong>Health endpoints</strong>: All five MCP servers expose a <code>health</code> tool returning server name, version, status, and timestamp -- addressing peer-review recommendation 1.7 (operational robustness).</li> <li><strong>Pagination guard</strong>: FHIR search now enforces <code>max_results</code> to prevent unbounded result sets -- addressing peer-review recommendation 1.3 (query guardrails).</li> <li><strong>Dataset governance</strong>: <code>datasets/README.md</code> with data dictionary, provenance, and known limitations; <code>datasets/manifest.json</code> with SHA-256 checksums -- addressing peer-review recommendation 3.1.</li> <li><strong>Repository-level files</strong>: <code>SECURITY.md</code>, <code>CONTRIBUTING.md</code>, <code>CITATION.cff</code> for publication readiness -- addressing peer-review repository-level recommendations.</li> <li><strong>3 new Mermaid diagrams</strong> in README: peer-review development lifecycle, quality assurance cycle, and MCP PAI oncology trial lifecycle.</li> <li><strong>Expanded test suite</strong>: 9 new tests including token expiry, encoded SSRF variants, health endpoints, policy traces, and cross-server integration trace.</li> <li><strong>Contributor icons</strong>: @kevinkawchak, @claude, and @codex profile images in README Team section.</li> </ul> <h2>Contributors</h2> <p>@kevinkawchak @claude @codex</p> <h2>Notes</h2> <ul> <li>All 39 tests pass. Test suite expanded from 30 (v0.1.0) to 39.</li> <li>All five MCP server versions updated to 0.2.0.</li> <li>This release addresses peer-review recommendations from <code>peer-review/trialmcp_pack_v0.1.1_peer_review.md</code> sections 1.1 (shared contracts), 1.2 (token lifecycle, policy traces), 1.3 (pagination), 1.4 (DICOM security), 1.7 (health endpoints), 3.1 (dataset governance), and repository-level recommendations (SECURITY.md, CONTRIBUTING.md, CITATION.cff).</li> <li>Remaining peer-review recommendations (1.3 de-identification profiles, 1.5 signature support, 1.6 W3C PROV alignment, 2.1 client state machine, 4.1-4.3 adversarial tests, paper/ directory, CI/CD workflows) are documented as future work.</li> <li>Python 3.11+ required. No new external dependencies.</li> </ul>
title TrialMCP Pack: MCP Servers for Physical AI Oncology Clinical Trial Systems
topic model-context-protocol
oncology
clinical-trials
physical-ai
robotics
fhir
dicom
audit
hipaa
mcp
url https://doi.org/10.5281/zenodo.18870616