Medication reconciliation and safety MCP server for the Agents Assemble hackathon. Provides six composable tools for healthcare agents: medication listing, drug-interaction analysis, renal dose checking, home-vs-hospital reconciliation, high-risk pattern screening, and reconciliation note generation. All outputs are FHIR-cited and require clinician review.
Argus demonstrates solid definition quality with all 6 tools having clear, descriptive names and documented input schemas. Tool names follow verb_noun convention (get_medications, check_interactions, renal_check, reconcile, screen_patterns, generate_note). All tools have descriptions (194-250 chars, within baseline 34-392 range). Input parameters are typed with descriptions. However, output schemas are not documented in the visible code, only input schemas are defined. Error handling and recovery guidance are absent from tool definitions. Tool composition is sound (each does one thing), but descriptions lack dependency hints and prerequisites that would optimize LLM planning.
Analyze drug-drug interactions (DDI) between medications with severity scoring and contextual risk factors
Generate a structured medication reconciliation note summarizing findings from all other tools with clinical recommendations
Retrieve active medications for a patient from FHIR MedicationRequest and MedicationStatement resources
Reconcile medications between home (patient-reported) and hospital (active orders) lists to identify discrepancies and intentionality
Verify medication dosing appropriateness based on renal function (eGFR) and identify drugs requiring dose adjustment or avoidance
Screen for high-risk medication patterns including Beers Criteria PIMs, anticholinergic burden, QTc-prolonging drugs, serotonin syndrome risk, and pregnancy category concerns
Output schemas not documented. Tool descriptions state what is returned (e.g., 'Retrieve active medications') but do not specify the structure of the response object, field names, or types. LLMs cannot plan downstream tool calls or extract data without knowing response shape.
No error handling or recovery guidance in tool definitions. Descriptions do not explain what errors might occur, how to interpret them, or what the LLM should do next (retry, ask user, try alternative tool). This violates pattern:recovery-guide.
Missing dependency hints and prerequisites. Descriptions do not explain when to call each tool or in what order. E.g., generate_note depends on outputs from check_interactions, renal_check, and screen_patterns, but this is not documented in the tool descriptions.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Parameter 'severity_threshold' in check_interactions uses enum constraint (minor|moderate|major) but description does not explain what each level means or when to use each. LLMs may select wrong threshold without guidance.
reconcile tool accepts both RxNorm codes and medication names in arrays but does not document the expected format, how to distinguish them, or what happens if both are mixed. This ambiguity invites LLM errors.