Posos MCP Server

What is MCP?

The Model Context Protocol (MCP) is an open protocol that allows AI assistants (such as Claude, ChatGPT, or any compatible agent) to use external tools in a standardized way. In practice, an MCP server exposes a set of tools that the AI assistant can automatically call whenever it needs to answer a question.

The Posos MCP server exposes all Posos medical APIs as MCP tools. This means your AI assistant can, in a single conversation:

• Search for a drug, allergy, or patient condition
• Analyze a prescription to detect interactions, contraindications, or dosage issues
• Look up the complete monograph for a drug
• Calculate clinical parameters (GFR, BMI, body surface area, etc.)
• Suggest therapeutic alternatives

All without the user needing to know the underlying APIs: the AI assistant selects and chains tools autonomously.

Getting connected

https://mcp.production.posos.co/.well-known/oauth-authorization-server

{
  "mcpServers": {
    "posos": {
      "url": "https://mcp.production.posos.co/api/v1/mcp"
    }
  }
}

curl -X POST https://zitadel.production.posos.co/oauth/v2/token \
  -d "grant_type=client_credentials" \
  -d "client_id=<YOUR_CLIENT_ID>" \
  -d "client_secret=<YOUR_CLIENT_SECRET>" \
  -d "scope=openid urn:zitadel:iam:org:project:id:333168797043131145:aud"

{
  "mcpServers": {
    "posos": {
      "url": "https://mcp.production.posos.co/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_TOKEN>"
      }
    }
  }
}

Available tools

The MCP server exposes 10 tools across 5 categories:

Search and identification

Drug autocomplete (drug-autocomplete)

Searches for a drug by name (even partial or approximate) and returns its identifiers.

The tool accepts a list of search terms and returns the best candidates for each, including:

• The Posos code (e.g., MV00001165) and associated terminology
• The full drug name (INN + dosage + form)
• The type: clinical drug (CLINICAL_DRUG), branded drug (BRANDED_DRUG), packaged drug (PACKAGED_DRUG), or ingredient (INGREDIENT)
• Ingredients and ATC classification

Condition autocomplete (terrain-autocomplete)

Searches for a patient condition (pathology, medical condition) and returns associated codes (SNOMED-CT, ICD-10). These codes are used to populate the patient profile when running contraindication analysis.

Allergy autocomplete (allergies-autocomplete)

Searches for an allergy by name and returns corresponding codes. Handles allergies to active ingredients, excipients, and drug classes.

Posology structuring (posology-structuration)

Converts free-text dosage instructions (e.g., “1 tablet morning and evening for 7 days”) into structured FHIR format. The result includes frequency, dose, and duration information in a format usable by the dosage analysis tool.

Each result includes a confidence score indicating the reliability of the structuring.

Medical database

Monograph (medical-database-getDrugMonographyByClinicalDrug)

Returns the complete monograph for a clinical drug: therapeutic indications, recommended dosages, adverse effects, alerts, and drug class interactions.

This tool expects a Posos clinical drug code (prefix MV, e.g., MV00000004). Use drug-autocomplete with the CLINICAL_DRUG type to obtain this code from a drug name.

Prescription analysis

These tools analyze a drug prescription taking the patient profile into account.

Contraindications and interactions (prescription-analysis-getContraindicationsAndInteractions)

Detects contraindications (related to the patient’s conditions) and drug interactions (between the drugs in the prescription).

Inputs:

• Drugs: list of drug codes (terminology posos, cis, or CIP13)
• Patient (optional): sex, date of birth, conditions (SNOMED-CT codes), biological observations (LOINC codes)
• Alert types (optional): filter by type (ABSOLUTE_CONTRAINDICATION, RELATIVE_CONTRAINDICATION, PRECAUTION, WARNING)

Cross-allergies (prescription-analysis-getAllergyIntolerance)

Detects cross-allergic reactions between the drugs in a prescription and the patient’s known allergies.

Inputs:

• Drugs: list of drug codes
• Allergies: list of allergy codes (SNOMED-CT or Posos drug classes, obtained via allergies-autocomplete)

Therapeutic alternatives (prescription-analysis-queryAlternatives)

Suggests safer alternative treatments for a given drug, based on a specific therapeutic indication and the patient profile.

Inputs:

• Drug to replace (code + terminology)
• Therapeutic indication (SNOMED-CT code)
• Filtering level: NO_RISK (no risk), NO_SEVERE_RISK (no severe risk), or ALL_ALTERNATIVES (all alternatives)
• Patient (optional): conditions and concurrent treatments

Dosage analysis

Dosage alerts (dosages-analysis-getDosageAlert)

Detects dosage issues for a list of drugs: overdose, underdose, excessive treatment duration, divisibility problems, etc.

Available alert types:

Medical calculators

Calculator (patient-calculators-GetCalc)

Performs common clinical calculations from patient data.

Available calculations:

Multiple formulas are available for some calculations (e.g., Cockroft, MDRD, CKD-EPI for GFR).

Codes and terminologies

The analysis tools use different coding systems. Here are the main ones:

Drug codes

Condition and observation codes

Common LOINC codes

Complete workflow example

Here is a typical conversation with an AI assistant connected to the Posos MCP server:

User: “My 72-year-old patient takes simvastatin and his doctor wants to add clarithromycin. Is there a risk?”

The AI assistant automatically chains the following tools:

This entire workflow happens transparently within a simple conversation, without the user needing to interact with the APIs directly.