Function Health MCP Server

The Function Health MCP server is the official remote connector that exposes a Function member's lab and nutrition data to AI assistants over the Model Context Protocol. It is hosted by Function Health at https://services.functionhealth.com/ai-chat/mcp and speaks MCP over Streamable HTTP, with authentication handled through Auth0 using OAuth 2.0 (with PKCE) and the MCP Authorization Spec dated 2025-06-18.

The connector is intentionally narrow in scope. It exposes three read-only tools: an overall lab results summary, a category-level breakdown across the 24 Function health areas (Heart, Metabolic, Thyroid, Liver, Kidney, etc.), and a personalized nutrition plan with goals and meal recommendations. Access is gated by OAuth scopes (read:health_summary, read:biomarkers, read:action_plan), and individual biomarker values, IDs, and personal identifiers are never shared with the model.

It is offered as a first-party Claude Connector and is available to Claude Pro and Claude Max users in the United States who also have an active Function Health membership with completed lab results. There is no public GitHub repository for this server: it is a provider-hosted remote MCP endpoint, not a self-hosted package.

Tools

Function Health hosts the MCP server. There is nothing to install locally.

Prerequisites

• Active Function Health membership with completed lab results
• Claude Pro or Claude Max subscription
• US-based account (the connector is limited to United States users)
• Initial setup must be performed on Claude.ai web or the Claude Desktop app (not mobile)

Connecting in Claude

1. Sign in to Claude.ai or open the Claude Desktop app
2. Go to Account Profile then Settings then Connectors
3. Open the Connector directory and select "Function Health"
4. Click "Connect" and complete the Auth0 / Function Health OAuth login
5. Approve the requested scopes (read:health_summary, read:biomarkers, read:action_plan)

Manual MCP client configuration

For MCP clients that support remote Streamable HTTP servers with OAuth, you can point the client at the Function endpoint:

{
  "mcpServers": {
    "function-health": {
      "url": "https://services.functionhealth.com/ai-chat/mcp",
      "transport": "http"
    }
  }
}

The server implements the MCP Authorization Spec (2025-06-18) and RFC 9728 Protected Resource Metadata, so a compliant client will discover the Auth0 authorization server and walk the user through the OAuth 2.0 + PKCE flow automatically. No API key is needed.

• Ask Claude for a plain-English summary of your latest Function lab panel without opening the Function app
• Identify which of the 24 Function health categories (Heart, Metabolic, Thyroid, etc.) currently have the most out-of-range biomarkers
• Generate meal ideas or grocery lists that align with the nutrition plan Function created from your lab data
• Compare in-range vs out-of-range counts over time and ask Claude to explain what changed
• Plan questions for a Function clinician visit by having Claude highlight categories that need attention

• "Summarize my overall Function lab results."
• "Which health categories are most out of range right now?"
• "Show me only the out-of-range biomarkers grouped by category."
• "What does my Function nutrition plan recommend, and can you turn it into a 5-day meal plan?"
• "How have my lab results changed since my last test?"

• Official, first-party connector built and hosted by Function Health, listed in Claude's Connector directory
• Strong security posture: OAuth 2.0 with PKCE, scope-based access, TLS in transit, and read-only tools
• Privacy-preserving by design: individual lab values, biomarker IDs, and PII are never sent to the model
• Zero install: works as a remote MCP server, no local package or API key management

• Read-only and summary-level: the model cannot see individual biomarker values, only counts and category-level data
• Limited to Claude Pro or Max users in the United States with an active Function Health membership
• Only three tools exposed, so deeper analytical use cases (per-test trends, raw values) are not possible through this connector

• Apple Health / HealthKit community MCP servers for self-tracked vitals and activity data
• Generic FHIR MCP servers (community-built) for connecting to electronic health record data
• Direct use of the Function Health web and mobile apps for full-fidelity biomarker values and trends