Screenshots

Cherry blossom map	Spot detail & weather

Fruit picking	What's On in Japan

Flowers in season	Mobile

The problem

You ask Claude, ChatGPT, or Gemini: "When should I visit Kyoto for cherry blossoms?"

You get "late March to early April" — the same answer every year, based on training data that has no idea what this winter was like.

The actual bloom date shifts by 2–3 weeks depending on temperatures. Tokyo peaked March 29 in 2024, March 22 in 2023, March 28 in 2022. Your AI is guessing. This MCP server stops the guessing.

The fix

It connects Claude (or any MCP client) to live forecast data from the Japan Meteorological Corporation, the same source behind SAKURA NAVI, Japan's most-used cherry blossom app. Data updates daily at 9AM JST.

You:    "I'm in Japan April 8–14. Where should I see cherry blossoms?"

Claude: Based on today's JMC forecast data:

        Cities in full bloom during your visit:
        • Sendai (Miyagi) — full bloom forecast Apr 10 (avg Apr 12)
        • Morioka (Iwate) — full bloom forecast Apr 13 (avg Apr 16, running early)
        • Akita — full bloom forecast Apr 14

        For Sendai, top spots include Tsutsujigaoka Park (92% full bloom today)
        and Nishi Park. Morioka Castle ruins are a classic.

        One thing to watch: 40% rain chance Apr 10 in Sendai. Rain speeds up
        petal fall, so Apr 11–12 is probably the safer window.

Real bloom percentages. Real park names. Real weather.

Install

Claude Desktop / Claude Code / any stdio MCP client

{
  "mcpServers": {
    "japan-seasons": {
      "command": "npx",
      "args": ["japan-seasons-mcp"]
    }
  }
}

HTTP endpoint — no install required

Any MCP client that supports HTTP transport can point directly at the hosted instance:

https://seasons.kooexperience.com/mcp

Self-host

PORT=3000 npx japan-seasons-mcp --http
# MCP endpoint: http://localhost:3000/mcp

What's covered

1,700+ GPS-tagged spots across 12 seasonal categories.

12 tools

Cherry blossom

get_sakura_forecast — the big picture

All 48 JMA observation cities in one call: this year's forecast, actual dates when observed, and how each city compares to the historical average. Good starting point before you drill into specific spots.

"What's the cherry blossom situation in Japan right now?"
→ 48 cities by region, bloom status, forecast dates, days vs average

get_sakura_spots — specific parks and temples

1,012 spots across Japan with bloom percentage, full-bloom percentage, and GPS coordinates, updated daily.

"Cherry blossom spots in Kyoto"
→ 51 spots: Kiyomizu-dera (87%), Maruyama Park (91%), Philosopher's Path (76%)...

get_sakura_best_dates — match travel dates to bloom

Give it your start and end dates, get back the cities where full bloom overlaps your window.

"I'm in Japan April 8–14, where should I go?"
→ Cities with bloom in that window, ranked by timing, avg comparison

get_kawazu_cherry — early-season deep-pink variety

Kawazu cherry blooms January–February in Izu Peninsula, months before standard sakura opens anywhere.

"Can I see cherry blossoms in February?"
→ 9 Kawazu spots with bloom %, GPS, forecast dates

Autumn leaves

get_koyo_forecast — maple and ginkgo timing by city

50+ cities with this year's colour-change dates and how they compare to the historical normal. Maple and ginkgo peak at different times; both are included.

"When do autumn leaves peak in Kyoto vs Hokkaido?"
→ City-by-city maple/ginkgo dates, days early or late vs average

get_koyo_best_dates — same idea as sakura best dates, for autumn

Match your travel window to cities in peak colour.

"I'm in Japan late October, where for autumn leaves?"
→ Cities in peak colour during your dates, maple vs ginkgo timing

get_koyo_spots — 687 viewing spots by prefecture

Each spot has a peak window (start, peak, end), leaf type, popularity rating, and GPS.

"Top autumn leaves spots in Kyoto"
→ Arashiyama, Eikando, Tofukuji, Rurikoin... with star rating and exact peak dates

Flowers

get_flowers — 80 curated spots, 8 flower types, Jan through Oct

Filter by type, prefecture, or month. Each spot has an official URL and verified GPS.

Festivals and events

get_festivals — 52 major recurring events with official URLs and attendance figures

"Best fireworks festivals in Japan?"
→ Sumida River (900k), Nagaoka (1.1M), Omagari, PL Osaka, Miyajima...

"Festivals in Kyoto in October?"
→ Jidai Matsuri (Oct 22), Kurama Fire Festival, with booking tips

Filter by type (fireworks / matsuri / winter), month, and prefecture.

Fruit picking

get_fruit_seasons — full-year calendar for 14 fruits

Which fruits are in season and at peak for any given month, with best regions and notes.

"What fruit can I pick in September in Japan?"
→ Grape at peak (Yamanashi, Nagano), Pear at peak, Peach ending, Apple starting

get_fruit_farms — 350+ farms with GPS and booking links

Pass month= and it auto-filters to farms with something in season. Add region= to narrow further.

"Strawberry farms near Tokyo in April"
→ Farms in the Tokyo/Kanto area with strawberry in season, GPS + Jalan links

Weather

get_weather_forecast — 3-day JMA forecast for 51 cities

Temperature, rain probability by 6-hour window, and conditions. Worth checking because rain speeds up petal fall.

"Weather in Osaka this weekend?"
→ Min/max temp, rain % per 6-hour window, conditions

How it works

flowchart LR
    subgraph live["Live APIs (cached 1–6h)"]
        JMC["Japan Meteorological Corp\nsakura · koyo · kawazu\n1,700+ spots · daily 9AM JST"]
        JMA["Japan Met Agency\nweather · 51 cities\nhourly"]
    end

    subgraph static["Static datasets (loaded at startup)"]
        DATA["flowers.json — 80 spots\nfestivals.json — 52 events\nfruit-farms.json — 350+ farms"]
    end

    subgraph server["japan-seasons-mcp"]
        MCP["12 tools\n1 planning prompt\nstdio + HTTP transport"]
    end

    subgraph clients["MCP clients"]
        C1["Claude Desktop\n/ Claude Code"]
        C2["Cursor\n/ Windsurf"]
        C3["Any MCP\nclient"]
    end

    JMC -->|live fetch| MCP
    JMA -->|live fetch| MCP
    DATA -->|in-memory| MCP
    MCP -->|MCP protocol| C1
    MCP -->|MCP protocol| C2
    MCP -->|MCP protocol| C3

Static datasets load at startup and are served from memory with no disk I/O per request. Live JMC data is cached server-side (1–6h TTL). The all-spots payload is pre-gzipped at startup so repeat serving is essentially free.

Bloom scale reference

The official JMC scale used for all 1,012 sakura spots:

BLOOM RATE — progress toward first bloom (開花)
─────────────────────────────────────────────────
 0%      60%       85%      100%
 │  Bud   │ Swelling │ Opening │ <- First bloom!
 花芽〜つぼみ  膨らみ始め   開き始め    開花

FULL BLOOM RATE — progress toward mankai / 満開
─────────────────────────────────────────────────
 0%   20%   40%   70%   90%  100%
 │Open│ 30% │ 50% │ 70% │Full │ <- Mankai!
 開花  三分咲き 五分咲き 七分咲き 満開

Peak viewing is typically full bloom ± 3 days. Rain accelerates petal fall.

Web app

seasons.kooexperience.com is the interactive companion to this MCP server. It shows all the same data on a map — 1,012 sakura spots with lifecycle colours (orange bud, pink bloom, green ended), 687 koyo spots, 350+ fruit farms grouped by location, and 80 flower spots. There's also a "Plan My Trip" mode where you pick cities and see every seasonal activity near each one ranked by distance, and a "Near Me" button that finds spots within 30km of your GPS location.

Development

git clone https://github.com/haomingkoo/japan-seasons-mcp.git
cd japan-seasons-mcp
npm install
npm run build
npm start            # stdio MCP mode
npm run start:http   # HTTP mode, MCP at http://localhost:3000/mcp

TypeScript. No external database. No auth required.

Data sources

Contributing

PRs welcome, especially for flower spots, festival entries, and farm corrections. See CONTRIBUTING.md.

Formerly

Previously published as japan-sakura-koyo-mcp (deprecated). Use this package instead:

npx japan-seasons-mcp

License

MIT · Built by Haoming Koo