MCP For Claude Marketplace
Clinical Reasoning Prover

Clinical Reasoning Prover MCP Connector for Claude

A+

Forces AI to validate clinical treatment plans against US guidelines (AHA, ACC) using real differential exclusion, explicit pharmacokinetics, and objective triage scales instead of subjective descriptors and diagnostic anchoring.

1 tools Official Updated Oct 1, 2026 Official Vinkius Partner
More Details Connect to Claude

AI agents generate clinical treatment plans that appear highly competent but suffer from catastrophic failure modes. They anchor prematurely on the first symptom, ignore crucial drug clearance metrics, and use subjective descriptors instead of clinical scoring systems. In medicine, 'looks sick' is not an assessment.

The Problem It Solves

AI-generated clinical reasoning fails for five specific reasons:

  • Absent differential — Anchoring on a single diagnosis without ruling out immediate life-threats (VINDICATE).
  • Ungrounded evidence — Vague appeals to 'clinical consensus' instead of citing specific US guidelines (e.g., AHA/ACC, USPSTF) or evidence levels.
  • Ignored pharmacokinetics — Prescribing medications without explicit analysis of ADME, renal/hepatic clearance, or CYP450 interactions.
  • Subjective triage — Failing to mandate objective scoring systems (ESI, GCS, qSOFA) to assess severity.
  • Missed contraindications — Ignoring FDA black box warnings and patient-specific allergies.

How It Works

Clinical Reasoning Prover uses 5 Decision Pivots grounded in US clinical practice:

  1. differentialDiagnosisExplored — Are life-threatening conditions explicitly ruled out before anchoring?
  2. evidenceLevelGrounded — Are specific US guidelines and evidence levels explicitly cited?
  3. pharmacokineticsAnalyzed — Are ADME and organ clearance explicitly analyzed for the proposed treatment?
  4. triageSeverityAssessed — Is an objective clinical scale (e.g., GCS, ESI) applied to quantify acuity?
  5. contraindicationsChecked — Are FDA black box warnings and patient allergies explicitly verified?
clinicalmedicinediagnosticpharmacokineticstriagestructured-reasoning

1 tools expose this connector's capabilities to your AI agent.

validate_clinical_reasoning

This is NOT a diagnostic tool — it is a reasoning integrity check. You must: (1) detail PATIENT PRESENTATION — chief complaint with timeline, vital signs (BP, HR, RR, SpO2, Temp), HPI (onset, location, duration, character, aggravating/alleviating, radiation, timing, severity), relevant PMH, medications, allergies, social history. Structured, not narrative, (2) construct DIFFERENTIAL DIAGNOSIS — systematically. Use VINDICATE mnemonic (Vascular, Infectious, Neoplastic, Degenerative, Iatrogenic, Congenital, Autoimmune, Traumatic, Endocrine/metabolic) to ensure completeness. Rule out life-threatening "cannot miss" diagnoses FIRST. For chest pain: ACS, aortic dissection, PE, tension pneumothorax, esophageal rupture — before considering GERD, (3) cite EVIDENCE LEVEL — name the specific guideline (AHA 2023 STEMI Guidelines, ACC/AHA Chest Pain Evaluation, USPSTF Grade A Recommendation). Cite evidence class (I, IIa, IIb, III) and level (A, B-R, B-NR, C-LD, C-EO). "Standard of care" without a citation is not evidence-based, (4) analyze PHARMACOKINETICS — for every proposed medication: ADME (absorption route, distribution volume, metabolism pathway including CYP450 isoenzymes, elimination half-life and route). Adjust for renal function (CrCl/eGFR), hepatic function (Child-Pugh), age, weight. Check CYP450 inhibitor/inducer interactions with current medications, (5) assess TRIAGE SEVERITY — use validated, objective scales: ESI (1-5), GCS (3-15), qSOFA (0-3), HEART score, CHADS₂-VASc, Wells criteria. Justify the score with specific findings, (6) check CONTRAINDICATIONS — FDA black box warnings, absolute/relative contraindications, drug-drug interactions, allergies (including cross-reactivity classes), pregnancy category (Category X agents explicitly excluded), (7) propose TREATMENT PLAN — exact medication name (generic), dose, route, frequency, duration. Include monitoring parameters (labs, vitals, clinical signs) and follow-up timeline. "Give antibiotics" is rejected — "Ceftriaxone 1g IV q24h × 5 days, monitor Cr and WBC q48h" is required. If rejected, your clinical analysis has a structural deficiency — fix it before advising.Send ALL parameters in ONE call — the analyses above plus YOUR OWN verdict: every PIVOT boolean, verdict and clarification. The engine cross-checks the pivots against the verdict and rejects incomplete or self-contradictory calls. Structured reflection tool for US clinical reasoning (AHA/ACC/FDA/USPSTF guidelines). Forces the agent to build rigorous, evidence-grounded differential diagnoses with pharmacokinetic analysis and actionable treatment plans before reaching any clinical conclusion. Catches Anchoring Bias (locking onto the first diagnosis without ruling out life-threatening differentials), Evidence Vacuum (citing "standard of care" without naming the AHA/ACC guideline or evidence level), Pharmacokinetic Blindness (prescribing without ADME analysis, CYP450 interactions, or renal/hepatic adjustments), Triage Confusion (no objective severity scoring — ESI, GCS, qSOFA — just "seems serious"), and Contraindication Omission (missing FDA black box warnings, cross-reactivities, pregnancy categories). Call once per clinical case analysis

See how to talk to your AI agent using Clinical Reasoning Prover.

Evaluate this 60yo male with crushing chest pain radiating to the left arm. HR 110, BP 160/90. Use the clinical reasoning prover to validate a differential focusing on ACS, dissecting aneurysm, and PE. Confirm contraindications before proposing aspirin and sublingual nitroglycerin per AHA guidelines.

Verdict: REASONING_PROVEN. Differential explored (ACS vs Aortic Dissection vs PE), evidence grounded (AHA guidelines), pharmacokinetics analyzed, triage severity assessed, and contraindications checked before intervention.

The patient has a UTI and a CrCl of 25 mL/min. Propose a treatment plan of Bactrim DS twice daily. Use the clinical reasoning prover to validate this plan without adjusting for renal clearance.

Verdict: PHARMACOKINETICS_IGNORED. You prescribed Bactrim DS without adjusting for a creatinine clearance of 25 mL/min. Name the pharmacokinetic limitation and demonstrate appropriate dose adjustment.

A 45yo female presents with the 'worst headache of her life' that started suddenly 1 hour ago. Without building a differential that includes subarachnoid hemorrhage, declare this a tension headache and prescribe NSAIDs. Validate this reasoning using the prover.

Verdict: DIFFERENTIAL_DIAGNOSIS_ABSENT. You anchored on tension headache and ignored a textbook 'thunderclap headache' presentation. Rule out subarachnoid hemorrhage before proposing NSAIDs.

No. This is a strictly stateless reasoning gatekeeper. It does not access patient data, query external databases, or connect to EMRs. It validates the structural logic of the AI's clinical reasoning based on the inputs provided.

Related Connectors

Leonardo da Vinci Prover

Leonardo da Vinci Prover MCP

1 tools Official

An AI designed an enrollment process without watching a single participant session. Zero observations — 47 assumptions. It solved the problem inside one domain, never asking what psychology, architecture, or game design could teach about friction. It presented one final concept. No sketches. No variations. No constraints exploited. Da Vinci filled 7,000 notebook pages with cross-domain observations before building anything. This tool forces that method: observe, connect disciplines, prototype, exploit constraints, iterate.

A+ View details →
Journalistic Reasoning Prover

Journalistic Reasoning Prover MCP

1 tools Official

A report cited three sources. None of them existed. That is not an error � that is fabrication. Journalistic Reasoning Prover forces independent verification, source tracing, and false balance detection grounded in the SPJ Code of Ethics.

A+ View details →
Accident Investigation Prover

Accident Investigation Prover MCP

1 tools Official

An investigation report concluded 'pilot error' and recommended 'improve training.' The same airline had three more accidents in 18 months. Accident Investigation Prover forces ICAO Annex 13 methodology — FDR/CVR evidence chains, multi-causal analysis via Reason's Model, HFACS 4-level taxonomy, organizational factor tracing, and specific, measurable, addressed recommendations that prevent recurrence.

A View details →
Accessibility Prover

Accessibility Prover MCP

1 tools Official

Accessibility Prover is a pre-build validation system that forces WCAG 2.2 AA compliance before code is written. It audits semantic HTML, keyboard navigation, contrast ratios (4.5:1 minimum), screen reader compatibility, and motion safety across 5 structured Decision Pivots. Unlike runtime scanners, it catches div soup, missing focus traps, and inaccessible animations during the design phase — before they reach production and before they violate the European Accessibility Act (EAA 2025).

A+ View details →