Clinical Trial Matching (ToolUniverse Claude Skill)

A ToolUniverse agent skill that takes a patient’s molecular profile and clinical state and returns a ranked, evidence-tiered list of candidate trials — searching the U.S., EU, and UK registries rather than ClinicalTrials.gov alone.


Type	Claude Skill (one of ToolUniverse’s pre-built agent skills)
Supplier	Zitnik Lab, Harvard Medical School
Availability	GA — part of the ToolUniverse skills collection (`skills/tooluniverse-clinical-trial-matching/`)
Pricing	Free / OSS (Apache-2.0); wraps public sources (ClinicalTrials.gov, EU CTIS, ISRCTN, CIViC, Open Targets, ChEMBL, DrugBank, MyGene, EFO/OLS, PharmGKB, FDA labels)
Capabilities	Read-only — registry and knowledge-base queries; produces a shortlist, not a clinical decision
Verified	works · 2026-08-13
Security	cleared · 2026-08-13 — mims-harvard/ToolUniverse Apache-2.0, read-only public sources

How to install

This skill calls ToolUniverse tools, so the ToolUniverse MCP server must be installed first (see the ToolUniverse page). Simplest registration:

claude mcp add --transport stdio tooluniverse -- uvx tooluniverse

Then add the skills:

Claude Code — install the whole skill collection (the skill resolves as tooluniverse-clinical-trial-matching):
```
npx skills add mims-harvard/ToolUniverse
```
Manual / other agents — copy just this skill directory into your skills folder:
```
git clone https://github.com/mims-harvard/ToolUniverse
cp -r ToolUniverse/skills/tooluniverse-clinical-trial-matching ~/.claude/skills/
```
(replace ~/.claude/skills/ with your agent’s skills directory if you are not using Claude Code/Desktop.)

The skill sets disable-model-invocation: true upstream, so invoke it explicitly (e.g. ask Claude to “use the clinical-trial-matching skill”) rather than relying on automatic dispatch.

What it does

Runs a ten-step matching workflow:

Profile standardization — resolve the disease to an ontology ID (OpenTargets_get_disease_id_description_by_name, ols_search_efo_terms), parse each alteration into gene + variant, normalize gene symbols via MyGene_query_genes.
Broad discovery — search by disease, biomarker, and intervention across search_clinical_trials (ClinicalTrials.gov), CTIS_search_trials (EU/EEA), and ISRCTN_search_trials (UK/international).
Characterization — batch-retrieve eligibility criteria, conditions and interventions, locations, status and dates, and outcome measures, 10 trials per batch.
Molecular eligibility matching — parse free-text eligibility and score molecular fit.
Drug–biomarker alignment — the trial’s drugs, their mechanisms, and FDA approval status for the biomarker combination.
Evidence assessment — cross-reference FDA approvals, published results, CIViC, and PharmGKB.
Geography and feasibility — site locations, enrollment status, proximity.
Alternatives — basket trials and expanded-access programs when no direct match exists.
Scoring and ranking — a composite Trial Match Score.
Report synthesis — an executive summary with ranked trials and a completeness checklist.

Evidence tiers: T1 FDA-approved or guideline-recommended biomarker–drug combination; T2 Phase III data; T3 Phase I/II data; T4 computational or preclinical only.

Trial Match Score (0–100) is the sum of four components — molecular match (0–40: exact variant 40, gene-level 30, pathway-level 20, none 10, explicitly excluded 0), clinical eligibility (0–25: all criteria met 25, most 18, some 10, ineligible 0), evidence strength (0–20: FDA-approved 20, Phase III 15, Phase II 10, Phase I 5), and trial phase (0–10).

Primary use cases: genotype-driven trial search for a precision-oncology or rare-disease case, biomarker-driven trial selection, eligibility triage across US/EU/UK registries.

Notes

It is a reasoning layer over ToolUniverse; without the MCP server registered, the tool calls fail.

This produces a shortlist for a clinician to review, not a clinical recommendation. Eligibility is scored by parsing free-text criteria, which is exactly where registry entries are least structured — a “criteria met” call is a hypothesis to confirm against the trial’s own screening process and the treating team. Registry records also lag reality: enrollment status, site activation, and cohort closures change faster than the postings. An excluded molecular call (score 0) is the one component worth trusting most, because exclusion criteria are usually stated explicitly.

Covering three registries is the distinguishing feature — a ClinicalTrials.gov-only search silently misses EU and UK trials for the same indication.

Complements Precision Oncology (interpretation of the same molecular profile) and Clinical Trial Design (the sponsor-side counterpart). The underlying registries and knowledge bases are catalogued separately: ClinicalTrials.gov MCP, CIViC, Open Targets, ChEMBL, DrugBank, MyGene, ClinPGx / PharmGKB. ToolUniverse ships ~68 such skills; other workflows are catalogued separately.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.