01. Architecture
Exposing AccuRXy to third parties
Six governed layers in the core platform sit behind a single API gateway. From there, two routes reach different kinds of partner, and both inherit the same controls.
AccuRXy core platform
Evidence Retrieval Layer
Hybrid search, jurisdiction filters, version control
Medicines Data Store
Structured records, interactions, HPRA and EMA
Clinical Rules Engine
Validation rules, dose checks, consistency
Scenario Orchestration
Evidence pack build, AI generation, schema
Validation Pipeline
Entailment, medicines, internal consistency
Academic Audit Trail
Immutable evidence pack, review workflow
API Gateway
Authentication, rate limits, jurisdiction scope
Partner REST API
HTTPS, JSON, bearer token auth, versioned
- myDispense dispensing simulation
- University VLEs: Blackboard, Canvas, Moodle
- CPD platforms
- HSE, NHS and health system training portals
POST /v1/scenarios/generate · GET /v1/scenarios/{id} · GET /v1/scenarios/{id}/evidence · POST /v1/evidence/search
MCP gateway
Model Context Protocol, tools and resources, JSON
- Claude Desktop
- ChatGPT
- Partner AI agents
- Cursor
search_approved_evidence · get_medicine_record · validate_dose · check_interaction · get_citation_bundle
Governance is not bypassed. Every API and MCP request passes through the same jurisdiction filtering, evidence versioning, validation pipeline and audit logging as the AccuRXy application itself.
02. Commercial offering
Two routes, one governance layer
Tools for pharmacy education at scale, across platforms.
Partner REST API
For software platforms embedding AccuRXy scenarios
- HTTPS REST endpoints with JSON request and response
- Bearer token auth, scoped per partner organisation
- 8 endpoints covering generate, retrieve, search, validate and cite
- Structured scenario output with embedded evidence pack
- Jurisdiction enforced: Irish sources, Irish law
- Audit log embedded in every scenario response
MCP Gateway
For AI agents and models connecting to AccuRXy evidence directly
- MCP, the open standard for model tool calling
- 8 tools covering search, medicines, validation, citations and templates
- Resources: source URIs, medicine records, competencies
- Compatible with Claude Desktop, ChatGPT and partner AI agents
- The model calls tools within its own reasoning, no custom code
- Same governance layer as the REST API, fully auditable
Illustrative integration partners
REST API
myDispense
Governed scenarios in dispensing simulation
REST API
University VLEs
Embed AccuRXy content in Blackboard or Canvas
REST API
CPD providers
Governed scenarios for pharmacist CPD modules
MCP
Claude and GPT agents
AI assistants access evidence tools directly
03. Partner tiers
Access levels
Academic, commercial and enterprise tiers, all under a Partner Agreement that covers acceptable use, data governance and attribution.
| Academic Partner | Commercial Platform | Enterprise | |
|---|---|---|---|
| Scenario generation | Up to 500 per month | Commercial volume | Unlimited |
| Evidence search | Full access | Full access | Full access |
| Medicine validation | Included | Included | Included |
| MCP gateway | Research use | Included | Included |
| Custom jurisdiction | Not included | On request | Included |
| White-label output | Not included | Available | Included |
| SLA / uptime | Best effort | 99.5% | 99.9% plus dedicated |
| Pricing | Institution licence | Per scenario or SaaS | Negotiated |
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.
04. Appendix I
Technical reference
Draft endpoint and tool listing as presented in the commercial deck.
Partner REST API, 8 endpoints
| Method | Endpoint | Description |
|---|---|---|
| POST | /v1/scenarios/generate | Generate a governed scenario from a structured spec. Returns scenario, evidence pack and validation status. |
| GET | /v1/scenarios/{id} | Retrieve a stored scenario, including evidence pack, review status and audit record. |
| GET | /v1/scenarios/{id}/evidence | Return the full evidence pack: passages, medicine records, clinical rules, model version, validation results. |
| POST | /v1/evidence/search | Search the evidence platform. Returns ranked, jurisdiction-filtered passages with metadata. |
| GET | /v1/medicines/{name} | Structured medicine record: strengths, formulations, route, indications, interactions, CD status, PCRS scheme. |
| POST | /v1/medicines/validate | Validate product, strength, dose, route and patient profile. Returns pass, flag or fail. |
| POST | /v1/scenarios/validate | Submit a scenario draft for schema, evidence-entailment, medicines and consistency checks. |
| POST | /v1/citations/check | Check whether a clinical claim is supported by a cited passage. Returns supporting text and metadata. |
Called by application code over HTTPS with JSON. All requests are jurisdiction-scoped and audit logged.
MCP gateway, 8 tools
| Tool | Inputs | Outputs |
|---|---|---|
| search_approved_evidence | query, jurisdiction, topic, source_type | Ranked, jurisdiction-filtered passages with source ID, version and effective date. |
| get_medicine_record | medicine_name, jurisdiction | Strengths, formulations, route, indications, interactions, CD status, PCRS scheme. |
| retrieve_source_passage | source_id, section_ref, version | Passage text, source metadata, version number, effective date. |
| validate_dose | medicine, strength, dose, route, patient_profile, jurisdiction | Pass, flag or fail with reasoning against structured records and clinical rules. |
| check_interaction | medicines, jurisdiction | Interaction status, clinical severity, management guidance and source reference. |
| get_scenario_template | error_type, setting, learner_level, jurisdiction | Validated template with mandatory sections populated and objectives mapped to the PSI framework. |
| validate_scenario | scenario_draft as JSON | Schema, evidence-entailment, medicines validation and consistency results. |
| get_citation_bundle | scenario_id | Every claim mapped to a source passage with source ID, version, date and supporting text. |
Called by AI models within their own reasoning. No custom integration code is needed from the partner, and governance matches the REST API.
Full specification
Authentication, jurisdiction scoping, MCP resources, the difference between the two routes and a worked myDispense integration are set out in the partner API and MCP specification.
