Wapimaji Mcp

💧 WapiMaji MCP — Kenya Water & Drought Intelligence

[wapimaji-mcp Glama score]

Install: pip install wapimaji-mcp · Use with any MCP client.

A Kenya drought-response toolkit for AI agents: simulated county drought phases (no real NDMA data is integrated yet), SMS alerts through Africa's Talking that need explicit confirmation, and drought events published to the coordination bus.

[License: MIT] [Python 3.11+] [MCP]

What it does, and does not

• Simulated drought phases for the 47 counties. Values are generated from the county name so downstream flows can be built and tested. They are not observations and not NDMA data; every answer says "synthetic": true. Never use them for decisions.
• No real data source is integrated. With SANDBOX=false the status tools return an explicit "not implemented, status unknown" error. NDMA publishes drought phases as periodic bulletins; FEWS NET and IPC publish food-security classifications through APIs that were slow or unverified when checked on 2026-10-07. Integrating one is the next step.
• SMS alerts through Africa's Talking: simulated by default (sent: false); a live send needs SANDBOX=false and confirm_send=true.
• Coordination: publish a drought event so downstream MCP servers (insurance, agriculture advisory, health, county alerts) can react.

Tools

Install

pip install wapimaji-mcp# or:uvx wapimaji-mcp

Configure

{  "mcpServers": {    "wapimaji": {      "command": "uvx",      "args": ["wapimaji-mcp"],      "env": {        "AT_USERNAME": "your_username",        "AT_API_KEY": "your_at_key",        "SANDBOX": "true"      }    }  }}

Data sources

Runtime safety and data limits

SANDBOX=true (the default) uses simulated county drought values, including aggregate alerts. These values are not observations from NDMA. Live aggregate alerts are not implemented and return an explicit error with unknown status.

SMS calls in sandbox return sent: false without contacting Africa's Talking. Live SMS requires both SANDBOX=false and confirm_send=true, supplied only after the user explicitly confirms that send. Provider acceptance does not establish delivery to the recipient.

No real data source is integrated. Earlier versions listed NDMA, the Kenya Meteorological Department and FEWS NET here; the code never used them.

Related

• mpesa-mcp — M-Pesa + Africa's Talking MCP server
• WapiMaji — The Streamlit dashboard version
• gabrielmahia.github.io — Full civic portfolio

IP & Collaboration

© 2026 Gabriel Mahia · [email protected] License: MIT Not affiliated with NDMA or Africa's Talking.

Related datasets

Earlier versions said this server draws on NASA MODIS NDVI, NOAA CHIRPS, TRMM and FEWS NET; it does not read any of them. The Hugging Face datasets africa-open-climate-data and east-africa-agricultural-pd are separate resources.

Live data

• river_discharge_forecast: forecast daily river flow (m³/s) and the peak day at a point on a river, from Copernicus GloFAS via Open-Meteo (CC BY 4.0). Not an official flood warning.
• rainfall_forecast: daily rain (mm) and chance of rain for up to 16 days at any location, from Open-Meteo (CC BY 4.0; free API for non-commercial use).

Each live answer names its source, attribution and retrieval time; if a service does not answer, the tool returns an error instead of a number. The other tools are unchanged.

Part of the East Africa Coordination Stack

This MCP server is part of the Kenya coordination infrastructure. Connect it to africa-coord-bus — the coordination event bus that routes signals between domains automatically.

pip install africa-coord-bus

All servers: pypi.org/user/gmahia Live demo: coord-cascade-demo

Part of the East Africa coordination stack

• Install & run: pip install reli-cli && reli list — the MCP servers on the official MCP Registry under io.github.gabrielmahia
• Evaluate any model on Swahili agent tasks: kipimo · dataset · leaderboard
• Coordinate across servers: africa-coord-bus — offline-first event bus with a built-in Kenya routing table
• Datasets: huggingface.co/gmahia · Docs hub: nairobi-stack

Model-agnostic by design: closed APIs, open-weight models, and small distilled models are all first-class citizens.