Every drug, one API. CH · EU · US — 28 public sources across 13 jurisdictions, staged, schema-validated, diff-tracked. REST + GraphQL + MCP, with a watchtower that catches upstream schema drift before you ship.
API Surfaces
Same data, three transports. Pick the one your runtime speaks — REST for HTTP clients, GraphQL for one-round-trip dashboards, MCP for AI IDEs and agents.
GET /v1/drugsTypeahead, lookup, monographs, interactions, changes.
Trigram search across product names, ATC prefixes, and identifier resolution from NDC, CIP-13, GTIN, Pharmacode, RxCUI, UNII, dm+d AMPPID, and Swissmedic-No. ETag + Cache-Control on every stable resource. Cursor-paginated change feed for cache invalidation.
POST /v1/graphqlOne round-trip for drug + ATC + interactions.
Multi-region drug-catalogue queries are read-heavy and varied — "resolve NDC → drug, CIP-13 → drug, GTIN → drug, then ATC parents + interactions with my current med list" wants one query, not five. Persisted-query path on Growth+ for high-volume clients.
npx @medisuisse/mcp-serverNative tools for Claude, Cursor, Windsurf.
A thin TypeScript MCP server. Schema definitions and an HTTP shim only — no curated data lives in the npm tarball. Free tools: search, ATC lookup, schema docs. Gated: get_drug, check_interactions, validate_prescription.
The Watchtower
28 sources across 13 jurisdictions — FDA NDC, DailyMed, RxNorm, dm+d, CIMA, BDPM, AIFA, BAG-SL, AIPS, EPha, EMA SPOR and more — drift upstream without notice. Every consumer rediscovers the breakage individually, usually weeks late, usually after a customer complaint. We catch it at staging.
Per-source assertions across all 28 feeds — BAG-SL column checks, AIPS / dm+d / CIMA XSD, EPha / FDA NDC / DailyMed JSON-schema, oddb and RxNorm per-column regex. Failed rows land in a per-source quarantine; promotion is blocked above a 1% fail-rate.
A run flagged `schema_drift` keeps the last-good snapshot in production, fires PagerDuty + Slack, and surfaces `schema_state: "stale"` in every API response so customers know before they ship a broken refill.
Every promoted row writes a row to `changes` (before, after, source, occurred_at). Subscribe via webhook or `GET /v1/changes?since=...` — most-requested feature from EHR vendors.
Every response carries `_meta.last_synced_at` and `schema_state` per source. The previous good snapshot is still served, but flagged. Silent shipping of broken data is the failure mode this exists to prevent.
_meta{
"drug": { ... },
"_meta": {
"ndc": { "last_synced_at": "2026-05-30T04:00:00Z", "schema_state": "ok" },
"dmd": { "last_synced_at": "2026-05-29T02:00:00Z", "schema_state": "ok" },
"cima": { "last_synced_at": "2026-05-30T03:00:00Z", "schema_state": "ok" },
"bag_sl": { "last_synced_at": "2026-05-15T05:01:00Z", "schema_state": "ok" },
"aips": { "last_synced_at": "2026-05-26T05:00:00Z", "schema_state": "ok" },
"epha": { "last_synced_at": "2026-05-25T03:00:00Z",
"schema_state": "stale", "reason": "schema_drift_2026-05-29" }
}
}Platform
ATC normalisation, cross-jurisdictional identifier bridges (NDC, dm+d, CIP-13, GTIN, Pharmacode, PZN, AIC), reimbursement tiers, and indication codes — what HCI, Vidal, Rote Liste, BNF, and USP-NF each charge separately, unified in one JSON shape with sane DX.
NDC ↔ RxCUI ↔ UNII ↔ dm+d AMPPID ↔ CIP-13 ↔ PZN ↔ AIC ↔ GTIN ↔ Pharmacode ↔ Swissmedic-No ↔ ATC. One indexed read on `/v1/drugs/lookup` for every dispense flow, every jurisdiction.
EPha-style severity (contraindicated · severe · moderate · minor · info), mechanism, recommendation. Lexicographically-sorted pairs, cross-jurisdictional by construction.
Subscribe to "what changed across NDC, dm+d, CIMA, BAG-SL this week" via webhook on Growth+, or cursor-paginate `GET /v1/changes?since=...` yourself.
`drugs.payload` carries the raw upstream fields from every source. When a customer asks "why is this strength 200 mg/ml and not 0.2 g/ml?", we have the answer without re-fetching.
A customer who already licences a paid catalogue — HCI, Vidal, Rote Liste, ABDA, BNF, RxNorm-paid extensions, USP-NF, KNMP G-Standaard — can supply their own credentials. Per-tenant only, never cached cross-tenant. We never become a reseller.
`/drugs` and `/interactions/check` do not touch PHI and need no DPA. `/prescriptions/validate` accepts bucketed context only (age band, eGFR band, pregnancy) and routes through PHI Gateway.
Drop @medisuisse/mcp-server into your editor and the AI resolver becomes us — not a tab switch to a national compendium. Free tier covers search, ATC tree, and schema docs. Paid tier unlocks full drug records, interactions, and prescription validation.
{
"mcpServers": {
"drugcatalog": {
"command": "npx",
"args": ["-y", "@medisuisse/mcp-server"],
"env": {
"MEDISUISSE_API_KEY": "ms_live_…",
"PHI_GATEWAY_API_KEY": "phi_live_…"
}
}
}
}Positioning
HCI, Vidal, Rote Liste, BNF, USP-NF, and G-Standaard are entrenched in their own jurisdictions. We won't out-cover any of them locally in year one. We win on worldwide coverage, developer ergonomics, and price.
Our foundation is 28 public sources — FDA NDC, Orange Book, DailyMed, RxNorm, UNII, dm+d, CIMA, BDPM, AIFA, AIPS, BAG-SL, EPha, oddb, EMA SPOR, WHO ATC and more. HCI, Vidal, Rote Liste, ABDA, BNF, RxNorm-paid extensions, USP-NF, KNMP G-Standaard are explicitly excluded from the shared catalogue. Customers who already licence them can BYOL credentials per-tenant; we never resell.
Every team that touches multi-jurisdiction drug data today writes the same 28 jobs: NDC weekly pull, DailyMed daily, RxNorm monthly, dm+d weekly, CIMA, BDPM, BAG-SL, AIPS XML parse, EPha normalize, oddb identifier bridge, EMA SPOR, WHO ATC, plus 16 more. Person-quarters of work, indefinite maintenance, and you still miss the schema drift. We run it, you call the API.
Developer tier is free with 10k req/month. Starter at $99, Growth at $490 — generous on purpose. HCI, Vidal, Rote Liste, BNF, USP-NF, G-Standaard each charge tens of thousands per tenant per jurisdiction. The data we ship is mostly public; restricting access aggressively kills adoption, and adoption is the only durable defence against a global incumbent modernisation.