MCP Server

Connect Claude and other AI assistants directly to ClinicaLister's live trial, drug, device, and regulatory data through the Model Context Protocol — capabilities, clients, data contexts, tiers, and security.

What the MCP server is

The ClinicaLister MCP server exposes the same clinical-trial intelligence you see in the app as live, queryable tools for AI assistants. It speaks the Model Context Protocol (MCP) — an open standard for connecting AI models to external data — so an assistant like Claude can search trials, resolve diseases through the knowledge graph, pull linked drug and device records, read pricing and safety signals, and generate full reports against live data instead of a stale export.

Note: The MCP server is a live connection, not a download. Every tool call runs against the current database — the same data the platform refreshes continuously from ClinicalTrials.gov, FDA, EMA, and the other sources.

What you can do with it

The server publishes 74 tools across 19 domains, plus 58 report recipes written for specific roles. At a high level:

Search and scope trials — free-text plus structured filters, and descendant-aware disease search through the MONDO knowledge graph.
Resolve a disease to its full subtype set, see the competitive landscape, and walk molecule ↔ disease ↔ trial relationships.
Pull linked FDA/EMA drug records, devices, substances, and labels for the trials in scope.
Read pricing (NADAC, ASP, Medicare Part D, Open Payments, shortages), post-market safety signals (FAERS, MAUDE), and regulatory change history.
Discover investigators and site contacts across the US and EU registries.
Run a persona-tuned report recipe end to end — competitive density, IP-cliff, pipeline depth, safety profile, market sizing, and dozens more.

Tip: New to the tools? Skim the MCP Tools Reference for the full catalog, and Report Recipes & Personas for ready-made report workflows.

The three data contexts

Every tool call runs inside an active context that decides what data is in scope. You set it once per session (or pass it per call) and it persists.

Contexts and what each one sees.
Context	Who can use it	Scope
Personal	All tiers	The trials you personally follow, and the drugs, devices, and molecules linked to them.
Team	All tiers, if you're on a team	Your team's followed trials and everything linked to them.
Investigation	Investigator, Elite, Enterprise	The full database — every trial, with rate limits in place of a follow-list.

Note: Aggregate knowledge-graph and public reference data (disease search, competitive landscape shape, reference pricing) are available to every tier. Enumerating individual trials outside your followed set is what Investigation context unlocks.

Connect from an AI assistant

The server lives at mcp.clinicalister.com and supports the standard MCP transports.

1
claude.ai (web & desktop) — recommended
Add ClinicaLister as a custom connector pointing at the MCP endpoint. You sign in with your ClinicaLister account through a secure OAuth flow (including Google sign-in) — the same account you use for the app — and Claude gains your entitlements automatically.
2
Claude Desktop (local config)
Add the ClinicaLister server to your MCP client configuration using the standard remote-MCP bridge. You authenticate once; trusted devices keep the session alive across restarts.
3
Agent SDK / custom clients
Any MCP-compatible client can connect over Streamable HTTP or SSE. Programmatic clients authenticate with a ClinicaLister token scoped to your account and tier.

Tip: Your license tier and team membership travel with your sign-in, so the assistant sees exactly the data and tools you're entitled to — nothing more.

License tiers and rate limits

Tools are gated by your ClinicaLister license. Higher tiers raise the rate limits and unlock full-database (Investigation) access.

Per-tier request limits.
Tier	Per minute	Per day	Investigation access
Basic	20	500	—
Pro	40	1,500	—
Investigator	60	3,000	Yes
Elite	80	5,000	Yes
Enterprise	100	10,000	Yes

Note: “Biggest license wins” across your personal, team, and organization memberships — you always get the most generous tier available to you.

Security and data boundaries

The MCP server enforces the same isolation as the app, at the tool layer.

In Personal and Team context, every tool is scoped to followed trials — drug, device, pricing, and regulatory tools resolve only from trials you follow.
Only verified, high-confidence links are exposed; unconfirmed, low-confidence links are never returned.
Connections use OAuth 2.1 with PKCE; tokens are stored hashed and trusted devices use proof-of-possession binding.
Every connection is rate-limited and IP-logged for abuse detection, with a circuit breaker on repeated auth failures.
Responses that could be misread when empty carry explicit data-quality flags and recovery hints, so an assistant never silently infers “no data” from a scoping boundary.