Ask your AI assistant about the sun. It can ask Sunwise.

Sunwise ships an open Model Context Protocol server, so Claude, ChatGPT, Cursor and any other MCP-compatible assistant can pull the live UV index, check a Fitzpatrick skin type, and get honest unprotected and SPF 50 burn-time estimates from the same algorithm this site runs — plus log sun sessions to your profile once you sign in.

What your assistant can do

Every tool returns a plain-English summary plus structured JSON. Estimates are educational, not medical advice.
ToolWhat it doesSign-in
get_uv_indexLive UV index, level, today's peak hour and max, tomorrow's max and current weather for a location.No
check_skin_typeEstimates Fitzpatrick type I–VI from the standard 8-question self-assessment, with burn threshold (MED) and example burn times.No
get_burn_timeUnprotected and SPF 50 time-before-burning right now, at today's peak, and the safest start times for a 90-minute session.No
log_sun_sessionSaves a finished sun session to your profile — Sunwise reconstructs the UV dose from the day's actual UV curve.Yes
get_sun_sessionsYour saved session history: durations, dose as % of burn threshold, peak UV, sunscreen use.Yes

Two endpoints

The public endpoint needs no account and serves the three information tools. The signed-in endpoint serves everything, and connects to your Sunwise profile through a standard OAuth sign-in so sessions your assistant logs land in the same history as the live sun-session tracker.

Public (no account):   https://sunwise.org.au/api/mcp
Signed in (your profile): https://sunwise.org.au/api/mcp/user

Connect in one minute

Claude (web or desktop): Settings → Connectors → Add custom connector, then paste the endpoint URL. Claude Code:

claude mcp add --transport http sunwise https://sunwise.org.au/api/mcp/user

Cursor and most other clients accept a JSON entry:

{
  "mcpServers": {
    "sunwise": {
      "url": "https://sunwise.org.au/api/mcp/user"
    }
  }
}

On first use of the signed-in endpoint your browser opens the normal Sunwise sign-in — a 6-digit code to your email, no password — followed by an approval screen. Prefer not to connect a profile? Use the public URL instead and skip sign-in entirely.

Things worth asking

  • What's the UV index at Bondi Beach right now — do I need sunscreen?
  • I'm skin type II in Brisbane. How long have I got unprotected at today's peak?
  • When's the safest time for a 90-minute run in Melbourne tomorrow?
  • Work out my Fitzpatrick skin type, then log my hour at the beach from 11:00 with two sunscreen applications.
  • How much sun have I had this week?

Frequently asked questions

What is the Sunwise MCP server?
An open Model Context Protocol endpoint that gives AI assistants the same engine behind Sunwise: live UV index for ~16,000 Australian locations (or any coordinates worldwide), the standard Fitzpatrick skin-type self-assessment, and time-before-sunburn estimates with and without SPF 50.
Do I need a Sunwise account to use it?
No — UV lookups, skin-type checks and burn-time estimates work anonymously on the public endpoint. An account (free, email-only) is only needed for the signed-in endpoint, which lets your assistant log sun sessions to your profile and read your history back.
Which AI assistants can connect?
Any MCP-compatible client: Claude (web, desktop and Claude Code), ChatGPT developer mode, Cursor, Windsurf, and most agent frameworks. Point the client at the endpoint URL and it discovers the tools automatically.
How does signing in from an AI assistant work?
Through standard OAuth. When your assistant first uses the signed-in endpoint, your browser opens the normal Sunwise sign-in — a 6-digit code to your email, no password — then an approval screen showing exactly what the assistant may do. You can deny it, and you can disconnect it later from your assistant's settings.
What can a connected assistant see and do?
After you approve it: read your saved skin type, log finished sun sessions (place, times, sunscreen applications — Sunwise reconstructs the UV dose from the day's actual UV curve), and read your session history. It never sees your email password (there isn't one) and nothing is shared with anyone else.
How are the burn times calculated?
The same model as the burn-time chart: your skin type's minimal erythema dose divided by the erythemal dose rate of 1.5 × UV index per minute, integrated over the day's real UV curve for timelines and session planning. Estimates are educational, not medical advice.
Is the MCP server free?
Yes. Like the rest of Sunwise, both endpoints are free. UV data comes from Open-Meteo; be reasonable with request volumes.

For developers and agents

Both endpoints speak streamable HTTP (stateless JSON-RPC POST). Auth on the signed-in endpoint is OAuth 2.1 with PKCE — discovery starts from the RFC 9728 protected-resource metadata; Client ID Metadata Documents and Dynamic Client Registration are both supported. A machine-readable site overview lives at /llms.txt, and the model behind every number is documented on the burn-time chart.

Rather check yourself?

One photo estimates your skin type and today's burn time — free, private, no account needed.

Check my burn time