Clinicaltrials MCP server

ClinicalTrials.gov — Clinical Trial Registry

The NIH-operated public registry of clinical trials worldwide. Every interventional study (and most observational ones) registered with the FDA must be filed here. ~470,000 studies covering Phase 1–4, conditions across the medical spectrum, and post-market safety surveillance. Free, no auth.

Part of Pipeworx — an MCP gateway connecting AI agents to 1679+ live data sources.

Why this matters for AI agents

If your agent is answering anything about drug development, clinical research, or medical product safety, ClinicalTrials.gov is the source of truth for what's being studied. The data is structured: NCT IDs, phases, statuses, sponsors, conditions, interventions, primary/secondary outcomes, results.

Three core flows:

1. Search. "What's being studied for Alzheimer's?" → ct_search({query: "Alzheimer's disease"}) → studies matching, with NCT IDs, phase, status.

2. Specific study. "Tell me about this trial." → ctgetstudy({nct_id: "NCT04280705"}) → full structured record including protocol, results, references.

3. Sponsor / volume. "How many trials does Pfizer run?" → ctsponsortrials({sponsor: "Pfizer"}) → enumerated by phase and status. See What sponsor means — it is the trials Pfizer leads, not every trial its drugs appear in.

4. Sponsor comparison. "Who has more recruiting Phase 3 obesity trials, Novo Nordisk or Eli Lilly?" → ctcomparesponsors({sponsors: ["Novo Nordisk", "Eli Lilly and Company"], condition: "obesity", status: "RECRUITING", phase: "PHASE3"}) → lead-sponsor counts ranked under identical filters with sample NCT records.

For drug-level synthesis (combining trials with FDA approvals and adverse events), use pharmadrugprofile compound or [compare_entities({type: "drug", values: [...]})].

Citable URI

pipeworx://clinicaltrials/study/{nct_id}

NCT IDs are stable forever. Once issued, never reused. Embed in agent output as the canonical study reference.

What sponsor means

Every tool that takes a sponsor matches the registered lead sponsor — the organisation that filed and runs the trial. That is what "what is Merck about to read out" means, and it is the default everywhere.

It is deliberately narrower than it looks. A university's study of pembrolizumab is not a Merck trial; Merck made the drug, someone else is running the study. To include those, pass sponsor_match:

Every returned study carries sponsormatchfield (leadsponsor or collaborator), so a caller can always see why* a given trial came back.

Until 2026-08-29 these tools filtered with the registry's query.spons parameter, which is not a sponsor filter at all — it is a broad search that also matches collaborators and free text. A Merck query returned 4,282 studies led by the University of Utah, Weill Cornell and others, with no indication in the response that anything was wrong. If you have cached counts from before that date, they are the broad number, not the lead-sponsor one.

Status filtering

When users ask about "active" trials, they usually mean one of two things:

• Currently recruiting: status=Recruiting — open to new participants
• In progress: status=Recruiting OR Active, not recruiting OR Enrolling by invitation

The default ct_search returns all statuses; filter on status field of results for the meaning you want. Common confusion: "Completed" trials are studies that finished collecting data, not necessarily ones with published results.

Phase transition rates (ctphasetransition_rates)

Shortened. The full README is on GitHub.