What MCP, A2A, and Google ADK Actually Unlock for African Coordination Infrastructure
Three protocols have matured enough to be meaningfully combined for African coordination work.
MCP (Model Context Protocol) — standard interface for AI agents to call external tools. A2A — agents coordinating with each other. ADK (Agent Development Kit) — building multi-agent systems.
Together they solve three distinct coordination problems.
The MCP problem
Most AI tools built for Africa are isolated. A model knows about M-PESA in general but cannot call the Daraja API. Knows about Kenya's counties but cannot query actual government data. Knows about drought patterns but cannot pull current satellite data.
MCP closes that gap with a standard interface between an AI assistant and an external service. There are now 31 MCP servers for East Africa on PyPI covering M-PESA, land records, agriculture, health, water, insurance, translation, county services.
The A2A problem
A single agent cannot handle all dimensions of an African coordination problem simultaneously. An agricultural advisor needs weather data, land records, market prices, credit history, and insurance eligibility — at once, from different systems.
A2A defines how agents communicate with each other. A coordinator delegates to specialists. The agricultural advisor delegates weather queries to a climate agent, land checks to an ardhi agent, market prices to a soko agent. Each does one thing well. The coordinator assembles the answer.
The ADK problem
The Agent Development Kit provides the application framework: agent lifecycles, tool registration, conversation management, deployment. Combined with MCP servers (real African services) and A2A (coordination protocol), ADK is the application layer that ties the stack together.
What this enables
A farmer in Murang'a asks about planting sorghum given the current water situation. The system checks NDVI drought indicators (wapimaji-mcp), queries historical planting calendars (kilimo-mcp), checks current market prices (soko-mcp), verifies water access rights (ardhi-mcp), checks parametric insurance products (bima-mcp), and synthesizes a recommendation in Swahili.
That coordination — across five systems, in real time, in the farmer's language — is what the combination of these protocols makes possible.
pip install mpesa-mcp wapimaji-mcp kilimo-mcp soko-mcp ardhi-mcp bima-mcp
The structural problem: African coordination requires combining multiple information sources simultaneously. These protocols exist for exactly that. The MCP servers for African services exist. The combination is infrastructure.
Top comments (0)