Kenyan diaspora engineer. Open-source coordination tools across insurance, credit, markets, civic access, health, and agriculture — built where information asymmetry costs the most. Swahili-native by default. AI is the delivery layer, not the headline.
M-PESA processes enormous transaction volume daily. Africa's Talking reaches users across 20+ countries on basic phones. Neither had an AI-agent interface. Drought signals, market prices, land records, and labor data each sit in their own silo — even when the response to one depends on knowing the other.
Each domain below represents a coordination problem — insurance, credit, markets, civic rights, health, agriculture. The tools are the infrastructure; AI agents are how a person interacts with them.
Each tool ships with CI, documented APIs, and a clear license. Built for someone to actually run, not as a demo.
Each tool below cites the paper or dataset its method comes from. The research existed; the regional implementation didn't.
Every tool is open-source, documented, and built to be extended. Start with mpesa-mcp and add the domain layer your project needs. Or take the whole stack in one step: pip install reli-cli && reli up. Every dataset maps to the server it backs and the eval that measures it in the data catalog; measure any agent against the kipimo leaderboard.
Concrete scenarios these tools are built to support.
TCP/IP didn't control the internet — it enabled it. These MCP servers are rails: structured, authenticated, locally-processable access to East Africa's institutional layer. The trains — health tools, civic assistants, agricultural bots, financial guides — are yours to build.
MIT licensed · pip installable