Documentation

v1.0
Live

MCP Server Documentation

Overview

The Ultrahuman MCP server lets you connect your own AI assistant — Claude, or any client that speaks the Model Context Protocol — to your Ultrahuman health data. Once connected, the assistant can call read-only tools covering sleep, recovery, HRV, movement, glucose, blood labs, Performance Labs and more, returning only your own data.

Key Properties

  • • Read-only: every tool is a read, nothing can be modified
  • • Per-user: tokens are bound to your account and can never return another user's data
  • • OAuth 2.1 with mandatory PKCE (S256) and user consent per scope
  • • Standard remote MCP server — works with any MCP-capable client
Server URL:https://api.ultrahuman.com/mcp

Limits & Behaviour

Rate limiting:cost-weighted per user; over budget returns HTTP 429 with Retry-After
Request body:max 64 KB
File access:reports and raw files are returned as signed, expiring URLs — never inline bytes
PKCE:S256 is mandatory; plain or missing code_challenge is rejected

You can review and disconnect connected assistants at any time — revoking a token immediately cuts off access.

Connect a Client

# claude.ai (web) or Claude Desktop
#
# 1. Open Settings → Connectors → "Add custom connector"
# 2. Name: Ultrahuman
#    URL:  https://api.ultrahuman.com/mcp
# 3. Click "Add" — you'll be redirected to the Ultrahuman
#    consent screen to sign in and approve the requested scopes.
#
# No client_id or secret needed: the client self-registers via
# OAuth dynamic client registration and completes the
# OAuth 2.1 + PKCE flow automatically.

Want to access your personal tokens and create OAuth apps? Login to view authenticated documentation →

API Terms