Research registered study records with AI and MCP

Run the workflow

1. 1. Choose a public research topic

   Call clinicaltrials_search_studies with a topic you intend to share with the public registry, a small limit and an optional source status filter. Do not provide personal health details. Results are topic candidates under the registry’s search semantics; matching is not proof of relevance, treatment suitability or individual eligibility.

2. 2. Inspect an exact NCT record

   Copy a returned NCT identifier into clinicaltrials_study_record. Review the exact canonical record, title, recorded sponsor, conditions and study type or phase. A validated missing ID is source absence at observation time. Errors and redirected aliases do not become missing records.

3. 3. Separate status, results availability and dates

   Retain overallStatus separately from hasResults: a completed study can have no posted results. Posted results do not establish efficacy, safety, completeness or a positive finding. Preserve month or day precision and ACTUAL or ESTIMATED milestones. Source processing date, record update date and ToolCargo observation time describe different events.

4. 4. Search publication candidates independently

   Use Europe PMC to find life-sciences publication records on the topic and inspect their source identifiers and publication types. Keep these candidates in a separate list. A similar title, condition or sponsor does not prove that a paper reports the registered study; verify any claimed relationship in the original records.

5. 5. Return a dated source-linked research list

   Attribute registry data to ClinicalTrials.gov and disclose the ToolCargo projection. Retain the actual source processing timestamp, API version, observation time, modifications and missing or truncated fields. Each operation reads fresh selected metadata and confirms matching source versions before and after the fetch. A detected source update returns an error and requires restarting. Explicit pages can still change, and separate backend requests are not an atomic snapshot. Review the live record again before later use.

What a useful result looks like

Two separate research lists: registry NCT ID → title → recorded study type/phase → overallStatus → posted-results availability → actual/estimated milestones and date precision → source processing date → source link → observation time → missing/truncated fields; publication source ID → citation metadata → independently reviewed relationship, if any.

Know the limits

ClinicalTrials.gov records are supplied by study sponsors and investigators; listing does not establish government review or approval of safety or scientific validity. Registry status is not treatment advice, individual eligibility or proof of efficacy. This connector reads selected public metadata, without outcome interpretation, participant information, contacts, narratives or files. Counts, rankings and pages can change. Source data retain international and possible third-party copyright; ToolCargo grants no reuse sublicense. One successful hosted operation consumes one shared-plan call, even though a source-version check, selected metadata fetch and confirming version check require three requests. No provider account, key or paid provider is needed.