| name | clinical-trials-search |
| description | Search ClinicalTrials.gov (API v2) for trials by condition, intervention, status, phase, location, or sponsor; fetch a trial by NCT ID; check eligibility; and count trials. Use for trial discovery, pipeline/portfolio analysis, or patient-matching context. Public API, no credential needed. |
| license | Apache-2.0 |
ClinicalTrials.gov Search (Codex-native)
Codex-native adaptation of Google DeepMind's clinical-trials-database skill
(science-skills, Apache-2.0).
Uses the public ClinicalTrials.gov REST API v2 directly.
Gates (ask before proceeding)
- Terms notice: on first use, tell the user to review the ClinicalTrials.gov
terms (https://clinicaltrials.gov/) and record the notice in provenance.
- Network / install: read-only public HTTP; be polite about rate limits. If
you write helper code, run it via
uv run, never bare python3.
Workflow
Search first with science_search_clinical_trials; use API v2 directly for
structured filters, pagination, field selection, and full records.
- Endpoint:
https://clinicaltrials.gov/api/v2/studies with query params —
query.cond (condition), query.intr (intervention), query.term,
query.spons (sponsor), filter.overallStatus (e.g. RECRUITING),
query.locn, and phase filters.
- Count first: request the total (
countTotal=true) before pulling records,
to gauge volume.
- Restrict fields: always pass
fields= (e.g.
NCTId,BriefTitle,OverallStatus,Phase,Conditions,InterventionName) — full
study JSON is very large. Write responses to a file and parse them.
- Paginate large sets with
pageSize + nextPageToken.
- Fetch one trial:
https://clinicaltrials.gov/api/v2/studies/<NCTId> for
full protocol, eligibility, outcomes, and locations.
- Report trials with NCT ID, title, status, phase, condition, intervention,
and sponsor; link each. Trust the server-side filters unless asked to verify
detailed eligibility.
- Provenance & review: record queries and cited NCT IDs with
$science-provenance; check claims with $science-review.
Boundaries
- Cite only NCT IDs actually returned; never invent NCT IDs, statuses, or
eligibility criteria.
- Trial records are registry data, not medical advice; do not infer efficacy,
safety, or eligibility decisions for an individual from them.
- Registration status and results reporting vary; state what the record does and
does not contain.