Technology & Governance

Partner API and MCP gateway

How AccuRXy makes its governed evidence platform, scenario generation engine and clinical validation capabilities available to authorised third-party platforms, AI systems and institutional partners.

Document
API and MCP Integration v1.0
REST endpoints
8, versioned and audit logged
MCP tools
8 tools plus versioned resources
Status
Working draft, 25 June 2026

01. Two integration models

One governed platform, two access routes

The two models serve different integration contexts and partner types, but both draw from the same governed evidence platform and produce academically credible, auditable outputs.

REST

Partner REST API

Designed for software platforms

A conventional HTTPS REST API that lets third party applications request governed scenarios, retrieve evidence, validate medicines and check citations programmatically.

Suitable for: Simulation platforms such as myDispense, LMS and VLE integrations, CPD providers and health service training systems.

MCP

MCP Gateway

Designed for AI agents and models

A Model Context Protocol server that exposes AccuRXy evidence and validation capabilities as tools and resources any MCP-compatible AI client can call.

Suitable for: Claude Desktop, ChatGPT integrations, partner AI agents, Cursor and future institutional AI assistants.

02. Partner REST API

Embed governed scenarios in your own platform

Third-party platforms authenticate with an API key, make structured requests and receive governed scenario content and evidence data in JSON.

Authentication and access control

ControlHow it works
AuthenticationAPI key issued per partner organisation, scoped to specific endpoint groups and passed as a Bearer token in the Authorization header.
Rate limitingPer-key limits enforced at the gateway. Academic research partners receive higher limits than commercial trial accounts.
Jurisdiction scopingEach key is bound to permitted jurisdictions, Ireland initially. Out-of-scope requests are rejected at the gateway before reaching the evidence layer.
Audit loggingEvery call is logged with partner ID, timestamp, endpoint, scenario specification and response hash. The log forms part of the immutable evidence pack.

Core endpoints

MethodEndpointDescription
POST/v1/scenarios/generateGenerate a governed scenario from a structured specification. Returns the scenario object with citations and validation status.
GET/v1/scenarios/{id}Retrieve a stored scenario by ID, including its complete evidence pack and audit record.
GET/v1/scenarios/{id}/evidenceReturn the full evidence pack: retrieved passages, medicine records, clinical rules, model version and validation results.
POST/v1/evidence/searchSearch the evidence platform directly. Returns ranked, jurisdiction-filtered source passages.
GET/v1/medicines/{name}Return structured medicine data for a product as licensed in the jurisdiction: strengths, formulations, routes, interactions, CD status, reimbursement.
POST/v1/medicines/validateValidate product, strength, dose, route and patient profile against structured records and clinical rules. Returns pass, flag or fail.
POST/v1/scenarios/validateSubmit a scenario draft for schema, evidence-entailment and internal consistency validation.
POST/v1/citations/checkSubmit a claim and a source ID. Returns whether the claim is entailed by the cited passage, with supporting text.

All endpoints return JSON. Scenario generation endpoints return a structured scenario object with an embedded evidence manifest.

Example request

POST /v1/scenarios/generate
Authorization: Bearer <partner_api_key>
Content-Type: application/json

{
  "jurisdiction": "ireland",
  "setting": "community_pharmacy",
  "learner_level": "fourth_year_mpharm",
  "topic": "anticoagulation",
  "error_type": "wrong_dose",
  "difficulty": "advanced",
  "learning_objectives": [
    "Identify a potentially unsafe warfarin dose",
    "Apply appropriate patient questioning",
    "Select the correct clinical intervention"
  ],
  "model_policy": "default",
  "output_format": "structured_json"
}

The response contains the scenario content, the evidence pack (source IDs, passages and medicine records used), validation results and an assurance summary. The partner platform renders this in its own interface and never needs to interact with the underlying evidence platform directly.

03. MCP gateway

Bring the evidence platform into an AI client

The Model Context Protocol is an open standard that lets AI models connect to external services through a consistent tool-calling interface. AccuRXy exposes its evidence and validation as tools and resources.

A pharmacy academic using Claude Desktop could connect to the AccuRXy MCP server and, without leaving their AI assistant, search approved evidence, retrieve the current Irish SmPC for a medicine, validate a dose, check an interaction or pull a pre-approved scenario template, all under the same evidence controls and audit trail as the AccuRXy application itself.

MCP tools exposed

ToolWhat it does
search_approved_evidenceSearch the curated, jurisdiction-filtered evidence platform. Ranked passages with source ID, version and effective date.
get_medicine_recordStructured medicine data: active ingredient, strengths, formulations, route, indications, contraindications, interactions, CD status, reimbursement.
retrieve_source_passageFetch a specific passage by source ID and section reference, with metadata and version, for citation verification.
validate_doseCheck whether a stated dose, strength, formulation and route is plausible for a described patient in the jurisdiction. Pass, flag or fail with reasoning.
check_interactionCheck for a clinically significant interaction across two or more named medicines, with severity, management guidance and source.
get_scenario_templateReturn a validated template for an error type, setting and learner level, with objectives mapped to the PSI competency framework.
validate_scenarioSubmit a draft for schema checks, evidence entailment, medicines validation and consistency flags.
get_citation_bundleReturn the full citation bundle: every claim mapped to a source passage with ID, version and date accessed.

MCP resources exposed

accurxy://sources/HPRA-104/version/3

accurxy://medicines/warfarin/ireland

accurxy://rules/high-risk-anticoagulants

accurxy://scenarios/templates/dispensing-error

accurxy://competencies/psi-ccf/v2

Resources are structured data items an AI model reads directly as context. They are identified by URI and are versioned.

How MCP differs from the REST API

Partner REST APIMCP Gateway
Who calls itSoftware applicationsAI language models within their own reasoning
How it is invokedExplicit HTTP request from application codeThe model decides when to call a tool, without the user needing to know it exists
Output handlingStructured JSON returned to the applicationStructured JSON returned to the model and incorporated into its response
GovernanceJurisdiction filtering, evidence versioning and audit loggingThe same governance layer, never bypassed
Primary partner typeSimulation, LMS, CPD and health service platformsAI agent platforms, research environments, academic tools

04. Commercial model

Partner tiers

Both access routes are offered under a tiered model, from academic research through to enterprise deployment.

Academic PartnerCommercial PlatformEnterprise
Scenario generationUp to 500 per monthCommercial volumeUnlimited
Evidence searchFull accessFull accessFull access
Medicine validationIncludedIncludedIncluded
MCP gatewayResearch useIncludedIncluded
Custom jurisdictionNot includedOn requestIncluded
White-label outputNot includedAvailableIncluded
SLA / uptimeBest effort99.5%99.9% plus dedicated
PricingInstitution licencePer scenario or SaaSNegotiated

All tiers are subject to a Partner Agreement covering acceptable use, data governance and attribution requirements. Jurisdiction scoping is enforced and every request is audit logged.

05. Illustrative use case

myDispense integration

A dispensing simulation platform gains governed, jurisdiction-specific clinical content without building an evidence infrastructure of its own.

myDispense is a web-based pharmacy dispensing simulation platform used in pharmacy schools across Ireland, Australia, the UK and beyond. It provides realistic dispensing environments but does not generate clinical scenarios with governed evidence packs or competency-mapped learning objectives.

myDispense provides the simulation environment. AccuRXy provides the governed clinical content. Neither platform needs to replicate the other's core capability.

  1. 01

    The myDispense administrator selects scenario parameters in their platform: learner year, error type and drug class.

  2. 02

    myDispense sends a POST request to /v1/scenarios/generate with those parameters and the Ireland jurisdiction flag.

  3. 03

    AccuRXy generates a governed scenario: retrieves Irish evidence, produces structured content, validates medicines and attaches citations.

  4. 04

    myDispense receives the structured scenario JSON and renders it inside its simulation interface.

  5. 05

    The student completes the dispensing task in myDispense, where their performance data is recorded.

  6. 06

    The evidence pack and assurance status remain available to the academic supervisor through the AccuRXy Evidence and Assurance panel.

From the student's perspective they are using myDispense. From the academic's perspective the clinical content is AccuRXy-governed and academically auditable.

06. Governance

What the API and MCP never bypass

Integration extends AccuRXy's reach without diluting academic integrity.

The Partner API and MCP gateway are access routes to AccuRXy governed capabilities. They do not bypass the governance layer. Every request, whether from the application, the REST API or an MCP tool call, passes through the same evidence retrieval, validation and audit pipeline.

Jurisdiction filtering

Applied to all API and MCP requests. A partner scoped to Ireland cannot retrieve UK-jurisdiction sources.

Evidence versioning

The evidence pack returned with any scenario records which source version was current at the time of generation.

Audit logging

Every call records what was requested, what evidence was retrieved, what the model produced and what validation results were returned.

Academic review status

A scenario generated via the API carries the same AI Draft status as one generated in the application. It cannot be marked Faculty Approved without human review.

Partner attribution

Scenarios generated via a partner API key carry a record of the requesting partner, preventing unauthorised reproduction or re-attribution.

Next

The same capabilities are packaged commercially through the API and MCP gateway. The gateway carries the architecture, partner routes and technical reference used in partner and investor discussions.