Floodwise MCP server

floodwise

England flood-risk by postcode for AI agents — verified Environment Agency data, not guesses.

Verified, trustworthy data tools for AI agents. "Qiniso" means "truth" in Zulu.

Website · npm · MCP Registry

Ask an LLM "what's the flood risk at SW1A 1AA?" and it will answer confidently — but it cannot know: per-postcode flood risk is specific, curated data that lives in a government dataset, not in a model's weights. floodwise looks the postcode up in the Environment Agency's official "Flood risk: postcode search tool" data and returns the real long-term risk band — or an honest "not found" instead of a guess.

⚠️ Read this first — what floodwise is and isn't

• England only. The data is the Environment Agency's, which covers England. Wales (Natural Resources Wales), Scotland (SEPA) and Northern Ireland (DfI) are not covered. A valid Welsh/Scottish postcode returns "not found", never a guess.
• Area-level, not property-level. The risk is for the area around the addresses in a postcode — not a specific building. The EA states this data is "generally not suitable for property level assessment".
• Long-term risk, not a live warning. It reflects the long-term annual chance of flooding, not whether flooding is happening now or forecast. It is not insurance, underwriting, surveying or professional advice.

Add it to Claude

floodwise runs locally over stdio (no key, no login). Once it's on npm, add it under mcpServers in your client config:

{ "command": "npx", "args": ["-y", "floodwise"] }

Use it as a library

npm i floodwise

import { floodRiskByPostcode, validatePostcode } from "floodwise";

floodRiskByPostcode("SW1A 1AA");
// {
//   postcode: "SW1A 1AA", found: true, headlineRisk: "Very Low",
//   addressesAtRisk: { high: 0, medium: 0, low: 0 }, groundwater: "Unlikely",
//   coverage: "England", dataset: "ea-official", ...
// }

floodRiskByPostcode("EH1 1AA").found;   // false — valid postcode, but Scotland (out of EA coverage), no guess
validatePostcode("ec1a1bb");            // { valid: true, postcode: "EC1A 1BB", outcode: "EC1A", incode: "1BB" }

Postcodes are accepted spaced or unspaced, any case. A well-formed postcode that isn't in the loaded England dataset returns found: false with a clear note — it never invents a risk level.

Tools — 2

Data

The flood data is the Environment Agency "Flood risk: postcode search tool data" (England), published as open data under the Open Government Licence v3.0. Each postcode carries the number of addresses whose surrounding area is at high (≥3.3%/yr), medium (1–3.3%) or low (0.1–1%) long-term risk from rivers, sea or surface water (the highest of these), plus a separate groundwater Possible/Unlikely indication. Refreshed roughly quarterly.

This repository ships an illustrative starter sample (non-geographic ZZ pseudo-postcodes) so tests run out of the box — every response from it is tagged dataset: "sample". To load the real data, download PostcodesRiskAssessment_All.csv from data.gov.uk / the Defra Data Services Platform and run:

bash

npm run build-data /path/to/PostcodesRiskAssessment_All.csv 2025-Q4

npm run build && npm test

Shortened. The full README is on GitHub.