# Clinical Reasoning Prover MCP > Clinical Reasoning Prover forces your AI agent to build rigorous clinical plans by checking them against specific US medical guidelines. It analyzes everything from life-threatening differentials to complex drug clearance metrics like CYP450 interactions, ensuring nothing critical is missed. ## Overview - **Category:** productivity - **Price:** Free - **Tags:** clinical, medicine, diagnostic, pharmacokinetics, triage, structured-reasoning ## Description When an AI generates a treatment plan, it often assumes too much or skips steps that are crucial in medicine. Standard reasoning fails because it anchors on the first symptom and ignores things like specific organ function or rare differentials. This MCP fixes that gap. It forces your agent to perform structured thinking by requiring explicit differential exclusion (using tools like VINDICATE), citing exact evidence levels from sources like AHA/ACC, analyzing how medications are metabolized in the body, and using objective triage scales instead of vague descriptions. If you connect this MCP through Vinkius, your AI client can validate these plans against a rigorous standard that demands more than just 'standard of care.' It provides accountability for every step: drug dosing, severity assessment, and contraindications. ## Tools ### validate_clinical_reasoning This tool rigorously checks a clinical case by confirming differential diagnosis, citing guideline evidence, analyzing drug metabolism, assessing severity, and verifying all contraindications. ## Prompt Examples **Prompt:** ``` 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. ``` **Response:** ``` 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. ``` **Prompt:** ``` 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. ``` **Response:** ``` 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. ``` **Prompt:** ``` 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. ``` **Response:** ``` Verdict: DIFFERENTIAL_DIAGNOSIS_ABSENT. You anchored on tension headache and ignored a textbook 'thunderclap headache' presentation. Rule out subarachnoid hemorrhage before proposing NSAIDs. ``` ## Capabilities ### Exhaustively rule out life-threatening diagnoses The system forces the agent to systematically check for immediate, high-risk conditions (like aortic dissection or pulmonary embolism) before settling on a primary diagnosis. ### Validate drug dosing based on organ function It analyzes patient kidney and liver function (CrCl/eGFR, Child-Pugh) to adjust medication dosages and check for dangerous metabolic interactions. ### Verify adherence to specific US guidelines The agent must cite the exact guideline (like AHA or USPSTF) and evidence class (I, IIa, etc.) supporting every recommended action or diagnosis. ## Use Cases ### Reviewing Chest Pain Workup A resident submits a chest pain protocol based on symptoms alone. The agent runs validate_clinical_reasoning, which immediately forces the inclusion of ACS and aortic dissection in the differential diagnosis before approving aspirin and nitroglycerin. ### Managing Renal Decline A patient needs antibiotics but has low kidney function (CrCl 25 mL/min). The agent runs validate_clinical_reasoning, which rejects the original dose and mandates a specific, adjusted drug regimen based on pharmacokinetic rules. ### Developing Protocols for Training A medical education tool uses this MCP to challenge trainee-generated plans. It requires them to cite evidence levels and use objective triage scoring before passing the protocol check. ### Handling Complex Comorbidities The AI needs to propose a drug, but the patient has multiple allergies and is pregnant. validate_clinical_reasoning checks for every single contraindication and flags the conflict immediately. ## Benefits - Reduces diagnostic omission. By running validate_clinical_reasoning, you guarantee the agent explicitly rules out life-threatening differentials before proposing a diagnosis. - Ensures accurate drug dosing. The MCP forces analysis of ADME and organ clearance (renal/hepatic), preventing dangerous medication errors that simple models miss. - Cites real evidence. It stops vague claims like 'standard of care' by demanding the specific guideline (e.g., AHA 2023) and its corresponding evidence level. - Mandates objective scoring. Instead of relying on subjective assessment, it forces the use of validated scales—ESI, GCS, qSOFA—to quantify patient acuity. - Flags drug conflicts. It checks for FDA black box warnings, allergies, and dangerous drug-drug interactions before confirming a treatment plan. ## How It Works The bottom line is that this MCP transforms an initial draft of clinical thinking into a highly scrutinized protocol, flagging structural gaps in logic and evidence. 1. You provide your AI client with a complete patient record, including chief complaints, vital signs, past medical history, and current medications. 2. Your agent runs the full analysis using validate_clinical_reasoning. This process forces checks across differential diagnosis, evidence citation, pharmacokinetics, triage scoring, and contraindications. 3. You receive a structured verdict indicating which critical steps were missed or if the proposed treatment plan is unsupported by US guidelines. ## Frequently Asked Questions **Can this MCP query patient records or EMR?** 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. **Why did the Prover reject my clinical plan with EVIDENCE_LEVEL_UNGROUNDED?** Because the reasoning relied on vague appeals like 'standard of care' or 'clinical consensus'. To pass the Prover, you must cite specific US guidelines (e.g., AHA/ACC, USPSTF, IDSA) or established evidence levels (e.g., Class I, Level A) to justify the intervention. **What objective scales are required for the Triage Severity pivot?** The Prover requires recognized objective scoring systems such as the Emergency Severity Index (ESI), Glasgow Coma Scale (GCS), qSOFA, or CHADS2-VASc. Subjective descriptors like 'very sick' or 'unstable' will trigger a TRIAGE_SEVERITY_BLIND rejection. **How do I best prepare my data before running `validate_clinical_reasoning`?** You must provide structured, comprehensive input covering the patient's full presentation and history. Focus on listing vital signs and chief complaints clearly rather than writing a narrative summary. **What happens if my proposed treatment plan conflicts with an FDA black box warning when I run `validate_clinical_reasoning`?** The MCP will immediately flag the conflict and refuse to validate the plan. It requires you to adjust the dosage or select an alternative medication that bypasses the contraindication. **Are there any rate limits or performance considerations when using `validate_clinical_reasoning`?** The MCP is designed for rigorous, deep analysis and has standard usage thresholds. For high-volume institutional use, please check the Vinkius dashboard for enterprise scaling options. **Is running through `validate_clinical_reasoning` compliant with HIPAA or PHI handling?** The MCP is a reasoning tool and does not store patient data; it processes temporary inputs. You must ensure your surrounding agent environment remains fully HIPAA-compliant. **If I cite evidence, but the guideline (e.g., AHA) has been updated since my source material, what will `validate_clinical_reasoning` do?** It flags the evidence level as potentially outdated and forces you to address the most current standard of care information available in its knowledge base.