Jma MCP Server

@pipeworx/jma

Japan Meteorological Agency (JMA) weather forecasts, warnings/advisories, and earthquake reports — keyless, sourced straight from JMA's public bosai JSON feeds.

Part of Pipeworx — an MCP gateway connecting AI agents to 1856+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

Tools

• jma_resolve_area(query) — resolve a place name (English romaji or Japanese, e.g. "Tokyo", "Sapporo", "大阪") or a JMA area code to the forecast OFFICE code used by jma_forecast / jma_warnings.
• jma_forecast(area) — 3-day short-term forecast (weather, precipitation probability, temperature) plus 7-day weekly forecast (weather, precipitation probability, forecast reliability, min/max temperature with confidence range), per sub-area. Accepts a place name directly — it resolves internally the same way as jma_resolve_area.
• jma_warnings(area) — current warnings and advisories (警報・注意報) for a region: headline bulletin plus per-area status (issued / continuing / lifted / none in effect) for every JMA warning category (heavy rain, storm, flood, wave, storm surge, dense fog, thunderstorm, avalanche, frost, and their "special warning" / emergency tiers), at both the sub-prefecture and municipality level.
• jma_earthquakes(limit?, min_magnitude?, min_intensity?) — recent earthquakes with magnitude, maximum seismic intensity (JMA's 10-step 震度 scale), epicenter, and report/occurrence time.

Auth

Keyless. No registration, no API key.

Data sources

• https://www.jma.go.jp/bosai/common/const/area.json — the area master list: 58 forecast offices plus every prefecture/region/city/ward/town JMA itself names, in both Japanese and English. This is what jma_resolve_area walks.
• https://www.jma.go.jp/bosai/forecast/data/forecast/<office>.json — the 3-day + weekly forecast for one office.
• https://www.jma.go.jp/bosai/warning/data/warning/<office>.json — current warnings/advisories for one office.
• https://www.jma.go.jp/bosai/quake/data/list.json — a rolling feed of recent earthquake bulletins (typically the last few weeks), newest first.

Traps for the next person

• Forecasts and warnings are published per forecast OFFICE, not per city. There are 58 offices, roughly one per prefecture plus Hokkaido and Okinawa subdivisions (e.g. Tokyo = 130000). A caller will say "Tokyo" or "Sapporo", not the office code — every tool that takes area resolves it through area.json's full hierarchy (office → region → city/ward/town), walking up to the owning office.
• Place-name ranking is (how well the name matches, then how broad the area is) — in that order, and the order is load-bearing. Both halves were silent wrong answers when they were the other way round: with level first, "Iga" (a city in Mie) returned a Niigata forecast because an office outranks a city no matter how badly it matches — same for Kaga→Kagawa, Oki→Okinawa, Aki→Nagasaki. With no match-quality signal at all, "Yokohama" returned Aomori, because JMA splits designated cities into sub-districts ("Northern" / "Southern Yokohama City", no plain 横浜市 leaf) and the only literal "Yokohama" leaf is a small town 700km away. The regression test that keeps this honest asserts all 2,217 English names in the area master resolve to themselves; run it before touching the ranking (fleet #1939).
• Same-named places in different prefectures are decided by area code (JIS prefecture order) and disclosed in alternatives. 伊達市 exists in both Hokkaido and Fukushima, 川崎町 in both Miyagi and Fukuoka — nothing in the query can separate them, so alternatives carries the others WITH the office each resolves to. alternatives: [] therefore means the name really was unique; it used to mean the opposite in exactly the cases that went wrong.
• area.json's four levels overlap in code space. A handful of small offices are their own single sub-region, so the same numeric code appears in both offices and class10s (e.g. 011000 Soya). The parent-walk in this pack always walks from the LEVEL you matched at (never by treating a code as globally unique across the four dicts), which is what makes that edge case resolve correctly instead of silently landing on the wrong office.
• weatherCodes is a closed, JMA-published numeric set (three digits, ~110 in active use) — this pack ships a full static translation table (weather_en) built from that fixed set, not a machine translation. The free-text weathers[] narrative (with time modifiers like "後" / "時々" / "一時") is intentionally left native-Japanese-only: it's compositional in a way a static table can't safely cover, and a wrong on-the-fly translation is worse than none.
• Warning codes (the code field inside each area's warnings[]) are a separate closed numeric set from weather codes (02-49, warnings / advisories / special warnings / emergency warnings). Cross-checked against a live Tokyo response: codes 14/15/16/20 in the areaTypes payload lined up exactly with "強風", "高波", "落雷", "濃霧" in that response's own headlineText, which is the validation this table was built against.
• Seismic intensity (maxi / int[].maxi) is JMA's own 10-step 震度 scale: 1,2,3,4,5-,5+,6-,6+,7 — 5-/5+ mean "5 Lower"/"5 Upper", not "slightly less/more than 5". An empty string means the bulletin carries no intensity (e.g. a "hypocenter update" record with no shaking report).
• area.json is ~260KB and essentially static (JMA's forecast-area boundaries are a standing administrative structure) — this pack caches it in module scope for 24h rather than refetching per call.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "jma": {
      "url": "https://gateway.pipeworx.io/jma/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/jma/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1856+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/jma_resolve_area \
  -H 'Content-Type: application/json' \
  -d '{"query":"Tokyo"}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/jma_resolve_area. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Run it from npm

@pipeworx/mcp-jma is a local stdio bridge to the Pipeworx gateway: your MCP client starts it with npx, and it forwards every call to the same https://gateway.pipeworx.io/jma/mcp endpoint shown above. It needs network access; there is no offline mode.

{
  "mcpServers": {
    "jma": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-jma"]
    }
  }
}

Without a key it calls anonymously, capped at 50 calls a day. A free account raises that to 200 a day: get a key at https://pipeworx.io/signup and add "env": { "PIPEWORX_API_KEY": "<your key>" } to the server entry. The key is sent as a Bearer token, the same way a hosted connection sends it, so calls are metered and billed exactly as they are there.

It lists only this pack's tools by default. Set PIPEWORX_META_TOOLS=1 to also list ask_pipeworx and the other gateway-wide tools.

Behind an HTTPS proxy: Node's built-in fetch does not read HTTPS_PROXY by default. Newer Node releases honour it when NODE_USE_ENV_PROXY=1 is set; check your Node version's documentation.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Jma data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides
• pipeworx.io

License

MIT