🛠️ CivicAgentKit — East African Civic AI SDK

Compatible with claude-sonnet-5 (released 2026-06-30) — Anthropic's most agentic Sonnet yet. Runs multi-step tool chains end-to-end without stopping short. Install: pip install civic-agent-kit · Use with any MCP client.

The unified Python toolkit for building civic AI tools in East Africa. One install gives you access to Kenya's parliament records, county budgets, embedded sample drought data (sources and as-of dates are not recorded), M-Pesa payments, and both MCP and A2A protocol integrations.

Install

pip install civic-agent-kit

What's included

from civic_agent_kit import (
    # Data loaders — Kenya's public civic datasets
    KenyaBudgetData,       # Controller of Budget — 47 counties
    KenyaParliamentData,   # MPs, bills, CDF — 13th Parliament
    KenyaSACCOData,        # SASRA SACCO registry
    KenyaDroughtData,      # drought phases (sample data; provenance not recorded)

    # Agents
    BudgetAgent,           # County budget analysis (CrewAI)
    RightsAgent,           # Constitutional rights Q&A (EN/SW)
    DroughtAgent,          # DEMO: synthetic drought values + SMS alert agent

    # Protocols
    KenyaMCPServer,        # MCP server — wraps all Kenya data as MCP tools
    KenyaA2AServer,        # A2A server — Kenya civic skills for agent-to-agent

    # Utils
    KenyaCounties,         # All 47 county codes + names
    KiswahiliTranslator,   # Simple civic term translations EN↔SW
)

Quick examples

Load county budget data:

from civic_agent_kit import KenyaBudgetData

budgets = KenyaBudgetData.load()
low_absorption = budgets[budgets['absorption_rate'] < 0.5]
print(low_absorption[['county', 'allocation_kes_m', 'absorption_rate']])

Serve Kenya civic data as an A2A agent:

from civic_agent_kit import KenyaA2AServer

server = KenyaA2AServer(host="http://localhost:8000")
server.run()  # exposes /.well-known/agent.json + JSON-RPC endpoint

Serve as an MCP server:

from civic_agent_kit import KenyaMCPServer

# Run alongside mpesa-mcp for full East African AI stack
server = KenyaMCPServer()
server.run()

Ask a rights question in Kiswahili:

from civic_agent_kit import RightsAgent

agent = RightsAgent(language="sw")
answer = agent.ask("Haki yangu ya ardhi ni nini?")
print(answer)
# → "Kifungu 40: Kila mtu ana haki ya kupata na kumiliki mali..."

The East African AI Stack

                    ┌─────────────────────────────────────┐
                    │         Your AI Agent                │
                    └─────────────┬───────────────────────┘
                                  │
               ┌──────────────────┴──────────────────────┐
               │                                          │
        ┌──────▼──────┐                          ┌───────▼──────┐
        │   MCP Layer  │                          │  A2A Layer   │
        │  (tools)     │                          │  (agents)    │
        └──────┬───────┘                          └───────┬──────┘
               │                                          │
     ┌─────────┴──────────┐                   ┌──────────┴─────────┐
     │                    │                   │                     │
┌────▼────┐        ┌──────▼─────┐     ┌───────▼────┐       ┌──────▼──────┐
│mpesa-mcp│        │KenyaMCP    │     │kenya-a2a   │       │CivicAgent   │
│payments │        │civic data  │     │parliament  │       │Kit agents   │
│SMS      │        │budgets     │     │budgets     │       │budget/rights│
│airtime  │        │parliament  │     │drought     │       │drought      │
└─────────┘        └────────────┘     │rights EN/SW│       └─────────────┘
                                      └────────────┘

Data

This package has two distinct data paths:

• File-backed loaders in src/civic_agent_kit/data.py. The county-budget loader points to the Kenya Civic Datasets Kaggle DOI 10.34740/kaggle/dsv/15473045; the broader dataset is also published on Hugging Face (DOI 10.57967/hf/8223).
• Embedded MCP demo tables in src/civic_agent_kit/server.py. Their original row-level sources and as-of dates were not recorded, so their drought, budget, parliament and SACCO values must be treated as synthetic/sample data, not official or current figures.

Dataset surfaces:

• Kaggle: gmahia/kenya-civic-data — DOI: 10.34740/kaggle/dsv/15473045
• HuggingFace: gmahia/kenya-civic-data — DOI: 10.57967/hf/8223

See Data provenance and truthfulness for the exact boundary between sourced files, unprovenanced embedded demo values, and constitutional-rights text.

Related packages

Package	Description	PyPI
mpesa-mcp	M-Pesa + Africa's Talking MCP server
civic-agent-kit	This package

Full portfolio

13 deployed civic apps: gabrielmahia.github.io

IP & Collaboration

© 2026 Gabriel Mahia · contact@aikungfu.dev License: MIT

Protocol Coverage — Complete Stack

civic-agent-kit now supports all four major AI agent protocols for East Africa:

Each protocol layer above was implemented within weeks of its global release — because the coordination gaps these tools address don't wait for tooling maturity.

from civic_agent_kit import CivicAgentKit

# Full protocol stack
kit = CivicAgentKit(
    mcp_servers=["mpesa-mcp", "wapimaji-mcp"],  # MCP
    a2a_registry="kenya-a2a",                    # A2A
    agui_server="kenya-agui",                    # AG-UI
    adk_config="kenya-adk",                      # Google ADK
)

Part of the East Africa coordination stack

• Install & run: pip install reli-cli && reli list — 33 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.