# Mould Detect Mycelium Network

> Remote MCP server and API for mould risk, building dampness, weather history, river flood and air quality, and mould-photo analysis. Free beta with a daily allowance. Australia-first; forecasts are global.

MCP endpoint: https://mcp.moulddetect.ai/mcp (Streamable HTTP; send `Authorization: Bearer <Mycelium key>`).
Get a key: https://mycelium.moulddetect.ai/keys/

## Docs

- [Getting started](https://mycelium.moulddetect.ai/docs/): connect Claude Code, Cursor, Codex or any MCP client
- [Tools reference](https://mycelium.moulddetect.ai/docs/#tools): inputs, coverage and cost of every tool
- [Data sources and licences](https://mycelium.moulddetect.ai/terms/#data-sources): attribution you must show with the data
- [Full text for LLMs](https://mycelium.moulddetect.ai/llms-full.txt)

## Tools

- `get_mould_risk_index` (Global forecast; history in Australia, 1 unit): Mould Growth Index (VTT model, 0 to 6) for a location from recent weather history and the 7-day forecast: today's index, the index in 7 days, trend, peak over 30 days and estimated days until visible growth, for a chosen surface material.
- `get_weather_forecast` (Global, 1 unit): Daily weather forecast up to 7 days (temperature, humidity, rain, wind) with optional mould-relevant variables such as humidity hours above 80%, condensation risk and drying index.
- `get_current_conditions` (Global, 1 unit): Latest weather conditions for a location plus today's high and low.
- `get_weather_history` (Australia only, 1 unit): Past daily weather from the ERA5-Land reanalysis: temperature, relative humidity, hours above 80% humidity, dew-point spread, condensation risk, rain and soil moisture. Useful to explain why a building is damp. Data ends about a week before today.
- `get_flood_risk` (Australia only, 1 unit): River-discharge forecast up to 30 days for the nearest river to a location, with the flood return period (2 to 100 years) each day meets. Returns an error if no river with at least 10 km² catchment is nearby.
- `get_air_quality` (Australia only, 1 unit): Air-quality forecast up to 5 days (PM2.5, PM10, ozone, nitrogen dioxide, sulphur dioxide, carbon monoxide, formaldehyde) and UV index, banded against WHO 2021 guidelines.
- `data_platform_status` (Global, 0 units): How current Mould Detect's data is: latest forecast run, latest complete day of weather history and latest air-quality run. Free to call.
- `analyze_mould_image` (Any photo, 20 units): Checks one photo of a wall, ceiling or surface for visible mould. Returns has_mould, no_mould or uncertain, plus mould_probability (0 to 1, calibrated). Uncertain means the photo needs a closer look or a professional inspection. Costs 20 units of the daily allowance.

## Optional

- [Mould Detect](https://moulddetect.ai/): the company and consumer app

## Connecting

Claude Code:

    claude mcp add --scope user --transport http mould-detect https://mcp.moulddetect.ai/mcp --header "Authorization: Bearer $MYCELIUM_KEY"

Cursor / Windsurf / Claude Desktop (mcp.json):

    {"mcpServers":{"mould-detect":{"url":"https://mcp.moulddetect.ai/mcp","headers":{"Authorization":"Bearer ${MYCELIUM_KEY}"}}}}

Codex (~/.codex/config.toml):

    [mcp_servers.mould-detect]
    url = "https://mcp.moulddetect.ai/mcp"
    bearer_token_env_var = "MYCELIUM_KEY"

Beta allowance: 500 units per day per account (an image analysis costs 20, other tools 1, data_platform_status 0), 60 requests a minute, 3 keys per account. Over the allowance a tool returns an error saying when it resets (00:00 UTC).

## Tool inputs

### get_mould_risk_index

Mould Growth Index (VTT model, 0 to 6) for a location from recent weather history and the 7-day forecast: today's index, the index in 7 days, trend, peak over 30 days and estimated days until visible growth, for a chosen surface material.

- `forecast_days` integer (optional): days of forecast to include, 1 to 7 (default 7)
- `history_days` integer (optional): days of past weather to include, 7 to 30 (default 30)
- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)
- `material_class` string (optional): surface sensitivity: very_sensitive, sensitive (default; timber, plasterboard paper), medium_resistant, resistant

### get_weather_forecast

Daily weather forecast up to 7 days (temperature, humidity, rain, wind) with optional mould-relevant variables such as humidity hours above 80%, condensation risk and drying index.

- `daily` string (optional): optional comma-separated extra daily variables, e.g. relative_humidity_hours_above_80,condensation_risk,mould_index,drying_index
- `forecast_days` integer (optional): days ahead, 1 to 7 (default 7)
- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)
- `past_days` integer (optional): recent past days to include, 0 to 7 (default 0)

### get_current_conditions

Latest weather conditions for a location plus today's high and low.

- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)

### get_weather_history

Past daily weather from the ERA5-Land reanalysis: temperature, relative humidity, hours above 80% humidity, dew-point spread, condensation risk, rain and soil moisture. Useful to explain why a building is damp. Data ends about a week before today.

- `end_date` string: last day, YYYY-MM-DD; clamped to the latest complete day (about a week ago); at most 366 days after start_date
- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)
- `start_date` string: first day, YYYY-MM-DD

### get_flood_risk

River-discharge forecast up to 30 days for the nearest river to a location, with the flood return period (2 to 100 years) each day meets. Returns an error if no river with at least 10 km² catchment is nearby.

- `days` integer (optional): forecast days, 1 to 30 (default 30)
- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)

### get_air_quality

Air-quality forecast up to 5 days (PM2.5, PM10, ozone, nitrogen dioxide, sulphur dioxide, carbon monoxide, formaldehyde) and UV index, banded against WHO 2021 guidelines.

- `days` integer (optional): forecast days, 1 to 5 (default 5)
- `latitude` number: latitude in decimal degrees, -90 to 90 (Sydney is -33.87)
- `longitude` number: longitude in decimal degrees, -180 to 180 (Sydney is 151.21)

### data_platform_status

How current Mould Detect's data is: latest forecast run, latest complete day of weather history and latest air-quality run. Free to call.

### analyze_mould_image

Checks one photo of a wall, ceiling or surface for visible mould. Returns has_mould, no_mould or uncertain, plus mould_probability (0 to 1, calibrated). Uncertain means the photo needs a closer look or a professional inspection. Costs 20 units of the daily allowance.

- `image_base64` string (optional): the photo as base64 (JPEG or PNG, up to 4 MiB before encoding; resize larger photos to about 1600 px or use image_url)
- `image_url` string (optional): public https URL of a JPEG or PNG photo, up to 8 MiB

## Every result

Results are JSON envelopes: `data` (the values), `attribution` (credits to show), `notices` (licence notices to pass on), `caveat` (limits to show), `freshness`, `coverage`, `terms`.

# Mould Detect data methodology

**Mould Growth Index** (`get_mould_risk_index`). The published VTT mould growth model (Hukka and Viitanen 1999;
parameterised by Ojanen et al. 2010), computed from 3-hourly temperature and relative humidity: up to 30 days of
ERA5-Land history (Australia) joined to up to 7 days of NOAA GFS forecast, for a material class (`very_sensitive`,
`sensitive` (default), `medium_resistant`, `resistant`). The index runs 0 to 6, labelled as in the API: 0 Dry, no
growth; 1 Starting (invisible); 2 Growing (invisible); 3 Visible spots; 4 Spreading; 5 Widespread; 6 Severe. `days_to_visible` estimates when the index reaches 3.
It is driven by outdoor conditions and does not measure indoor surfaces. Outside Australia the history part is
unavailable and the result is forecast-only (`history_available: false`).

**Weather.** Forecasts and current conditions: NOAA GFS, four runs a day, global. History: ERA5-Land reanalysis,
Australia, about 5 to 7 days behind real time (see `history_latest_complete_date` in `data_platform_status`).
Mould-relevant fields: relative humidity hours above 80%, dew-point spread, condensation risk, soil moisture.

**Flood.** Copernicus GloFAS river-discharge forecasts up to 30 days, Australia. The point snaps to the nearest
river cell with at least 10 km² upstream area; each day reports discharge (m³/s) and the highest return period
(2 to 100 years) it meets.

**Air quality.** Copernicus CAMS forecasts (PM2.5, PM10, ozone, NO₂, SO₂, CO, formaldehyde) and UV index up to 5
days, Australia, banded against WHO 2021 guidelines.

**Image analysis** (`analyze_mould_image`). A Mould Detect classifier (MobileNetV3) returns `has_mould`,
`no_mould` or `uncertain` with `mould_probability`, the calibrated probability (0 to 1) that the photo shows mould. `uncertain` is a deliberate answer: the photo needs a closer look or a
professional inspection. It cannot identify species or health risk.

All results are decision support only. They are not medical, laboratory or building-inspection results.

# Data sources and licence notices

Every Mycelium tool result includes `attribution`, `notices`, `caveat` and `terms`. Show them to your user with
the data and pass them on if you republish it.

| Source | Licence | Wording |
|---|---|---|
| ERA5-Land (history) | Licence to Use Copernicus Products | "Contains modified Copernicus Climate Change Service information {year}" |
| CAMS (air quality) | Licence to Use Copernicus Products | "Generated using / Contains modified Copernicus Atmosphere Monitoring Service information {year}" |
| GloFAS (flood) | CEMS-FLOODS datasets licence | "Generated using / Contains modified Copernicus Emergency Management Service information {year}" |
| NOAA GFS (forecasts) | US public domain | "NOAA Global Forecast System (GFS), public domain" |

The European Commission and ECMWF are not responsible for any use of this information. Data is provided as is,
for information only, without warranty. Full terms: https://mycelium.moulddetect.ai/terms#data-sources