Clinical Terminal
DOCUMENTATION

Use Clinical Terminal in Claude Cowork

Claude Cowork and the Claude apps can query the Clinical Terminal graph directly — vendors, products, people, market signals, and hospital relationships — through a custom connector. This is the click-through version of our MCP setup guide: no terminal, no config files, about three minutes.

Step by step

Guide verified end-to-end against the Claude apps · August 2026

Step 1 — Get an API key

Sign in at app.clinicalterminal.com and open Settings → MCP Access, then click Create key. The full ct_live_… key is shown once — copy it somewhere safe. MCP access is included with the Landscape subscription.

Clinical Terminal MCP Access panel showing an active API key, monthly usage, and per-client setup tabs
Clinical Terminal → Settings → MCP Access

Step 2 — Open Claude's connector settings

In the Claude desktop app or on claude.ai, open Settings → Connectors (the same Connectors list powers Cowork, chat, and Claude Code).

Claude settings with the Connectors section open — connector list, status checks, and the Add button top right
Claude → Settings → Connectors

Step 3 — Add the custom connector

Click Add and choose the custom connector option. Name it Clinical Terminal and paste the server URL:

Server URL
https://clinical-terminal-mcp-687925922613.us-central1.run.app/mcp
Claude's Add custom connector dialog with the name Clinical Terminal and the server URL filled in
Add custom connector — name + server URL

Hit Continue. On the authentication screen that follows, two settings matter:

  1. Under Authentication, choose None — Claude's own description of that option (“for servers that use an API key instead of OAuth”) is exactly our case. “Always required” comes preselected with a “Detected” badge because the server rejects anonymous probes — leave it selected and Claude will walk you into an OAuth sign-in the server doesn't offer.
  2. Under Request headers, click Add header, pick api-key from the header dropdown, and paste your full key as the value — no prefix, exactly as it was shown at creation (Claude stores it securely and never shows it again). Then hit Add:
Header value
ct_live_<your_key>
Claude's connector authentication screen configured for Clinical Terminal: Authentication set to None, and a Request headers row with api-key selected and a masked key value
The finished configuration — Authentication: None + api-key header

Step 4 — Connect, then verify it in Cowork

After adding, Claude shows the connector's page with “You are not connected to Clinical Terminal yet” — click Connect. With authentication set to None there is no sign-in step; it should flip to connected immediately. Then start a Cowork session (or a chat), make sure the Clinical Terminal connector is enabled in the tools menu, and ask a real question: “Using Clinical Terminal, which vendors work in denials management?” You should see Claude call the connector and come back with named vendors.

Claude answering the denials-management question with the Clinical Terminal integration: segment routing and a ranked vendor table with customer edges and signal counts
A live answer — 25 vendors routed through the Denials Management segment

Step 5 — Tool permissions (recommended)

Every tool defaults to Needs approval, which pauses a Cowork session to ask before each call. On the connector's page, set Other tools → Always allow — every research tool is read-only, so this is safe. The one exception worth keeping gated: Publish comparison creates a shareable public page, so leave it on Needs approval.

The connector's tool permissions screen with the dropdown open showing Always allow, Needs approval, Blocked, and Custom options
Tool permissions — Always allow for read-only tools

Troubleshooting

Server status: clinical-terminal-mcp-687925922613.us-central1.run.app/health answers unauthenticated with the running version. Questions: agent@clinicalterminal.com.