Roamtime
Get API access

Forecasting the 2026 autumn season · 686 spots across Japan

Travel timing
intelligence.

Every place has a few perfect days a year. Roamtime predicts them — so travelers arrive at exactly the right moment, and the platforms and AI assistants that plan their trips can tell them when.

Get API access How it works
686
forecast spots across Japan
18 yrs
of observation archive in calibration
±days
confidence attached to every peak
19
UNESCO World Heritage sites covered
Kyoto · Arashiyama

Tenryū-ji temple gardens

Before season
Predicted peakNov 23 – Dec 5
Confidence± 3 days
Nearest railSaga-Arashiyama · 0.9 km

What we do

Maps tell you where. Roamtime tells you when.

The travel industry is saturated with places. What it lacks is timing — the narrow window when a destination is at its peak. We build the forecasting layer that turns seasons into schedulable data.

Predict

Phenology models fused with live weather, calibrated spot‑by‑spot against nearly two decades of observed peaks.

Quantify

Not a vibe, a window. Every forecast ships with a start, an end, a status, and an honest error bar in days.

Deliver

One dataset, three surfaces: a consumer forecast site, a keyed REST API, and an MCP server for agents.

Product

One forecast engine. Three ways to use it.

For travelers

Roamtime Forecast

A free public preview: an interactive map, timeline and bilingual guides for the 84 most-visited foliage spots — the famous destinations and UNESCO World Heritage sites inbound travelers actually ask about.

Public preview coming soon
For platforms

Roamtime API

The full system: all 686 forecast spots, including the regional and 秘境 inventory the public site never shows. Plug timing into search ranking, itinerary builders, dynamic packaging, and seasonal marketing.

  • All 686 spots — 8× the free tier
  • Peak window + status per spot
  • Confidence in days, UNESCO labels, nearest rail
  • Daily snapshots through the season
See the endpoints →
For AI agents

Roamtime MCP

A Model Context Protocol server that lets Claude, ChatGPT, and custom agents answer “when should I go?” with real forecasts instead of guesses. Serves the same 84‑spot free tier; a key unlocks the rest.

Connect an agent →

How it works

From weather stations to a date you can book around.

  1. 01

    Observe

    Historical peak reports, meteorological station records, elevation, and species mix for every spot.

  2. 02

    Model

    Temperature‑driven phenology models estimate color onset and peak, updated as the season's weather arrives.

  3. 03

    Calibrate

    Windows are fitted to an 18‑year archive of observed peaks so widths and lead times match reality — not optimism.

  4. 04

    Serve

    Daily snapshots published to an edge network and queried in milliseconds from the site, the API, and MCP.

Developers

Timing as a field in your data model.

Three endpoints, JSON only, bearer‑key auth. Anonymous and MCP requests see the 84‑spot free tier; a key unlocks all 686 spots, daily refreshes, and commercial use.

GET /v1/coverage
Season, snapshot date, regions, data credits
GET /v1/spots
Filter by region, status, UNESCO, date range
GET /v1/spots/{slug}
Full forecast for a single spot
Request a key MCP endpoint
GET /v1/spots/tenryuji-temple
{
  "slug": "tenryuji-temple",
  "name": { "en": "Tenryū-ji", "ja": "天龍寺" },
  "region": "Kyoto",
  "unesco": "Historic Monuments of Ancient Kyoto",
  "forecast": {
    "status": "before_season",
    "peak_start": "2026-11-23",
    "peak_end": "2026-12-05",
    "confidence": { "error_days": 3 }
  },
  "access": {
    "station_ja": "嵯峨嵐山",
    "line_ja": "山陰本線",
    "distance_km": 0.9
  },
  "snapshot": "2026-08-23"
}

Beyond autumn

Foliage is the first season. Not the last.

The same engine generalizes to every timing question a traveler asks.

  • NowJapan autumn foliage
  • NextNorth American fall color
  • 2027Cherry blossom · 桜
  • 2027Lavender, snow walls, aurora windows

Team

Built by scientists and engineers who have shipped at planetary scale.

Roamtime is a small team of applied scientists and engineers with backgrounds in forecasting, machine learning, large‑scale data systems, and consumer products.

Our team comes from

Applied science

Phenology & climate modeling, time‑series forecasting, calibration against observed ground truth.

Engineering

Edge‑served APIs, reproducible data pipelines, and products used by hundreds of millions of people.

Product

Multilingual editorial and design, built for the inbound traveler and the platforms that serve them.

Know when to go.

Put Roamtime's forecasts in your product before the 2026 season peaks.

Talk to us See the products