io.github.gabrielmahia/sifa-mcp

⭐ sifa-mcp — Portable Reputation & Skills Passport MCP Server

---

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 sifa-mcp · Use with any MCP client.

---

Sifa = reputation/praise in Swahili.

Informal-sector workers build reputation with one buyer, one platform, one neighborhood at a time — and lose it the moment they switch. Nothing makes that reputation portable or agent-queryable.

The Structural Problem

In mature economies, reputation travels: Uber driver ratings, Yelp reviews, LinkedIn endorsements, contractor licence registries. A skilled plumber in Chicago can prove her track record to any homeowner in Los Angeles.

In Kenya, trust is tribal and local. A qualified mason in Mombasa cannot take her reputation to Nairobi. Economic mobility is throttled by the non-portability of professional identity.

Reputation portability is a coordination technology. Without it, hiring defaults to personal networks, which excludes the most talented people outside existing networks.

Tools

Tool	What it does
create_worker_profile	Create a portable reputation profile with skills, location, experience
add_service_record	Log a completed job with client details and rating
get_reputation_profile	Retrieve full profile with trust score for any requester type
generate_reference	Generate a formatted professional reference letter
list_skill_categories	Browse all NITA-aligned skill categories
compare_workers	Rank multiple workers by trust score and skill match

Trust Score

Scored 0–100 across four signals:

• Job volume (up to 30 pts) — number of completed jobs
• Average quality (up to 35 pts) — client star ratings
• Verified records (up to 20 pts) — client-OTP-confirmed completions
• Tenure (up to 15 pts) — years of documented history

Tiers: NEW → BUILDING → ESTABLISHED → TRUSTED

Quick Start

pip install sifa-mcp       # coming soon to PyPI
sifa-mcp                   # starts on stdio

Example

# Create a mason's portable profile
create_worker_profile(
    full_name="James Mwangi",
    phone="0712345678",
    primary_skill="construction",
    sub_skills="masonry,tiling,plastering",
    location="Nairobi",
    years_experience=8
)
# → SFA-A3F2B19C4E1D

# Log a completed job
add_service_record(
    profile_id="SFA-A3F2B19C4E1D",
    client_name="ABC Contractors",
    client_phone="0722000000",
    service_type="Kitchen tiling",
    duration_days=5,
    payment_kes=15000,
    client_rating=5
)

# Generate a reference for a new job application
generate_reference(profile_id="SFA-A3F2B19C4E1D", reference_purpose="job_application")

Research Basis

• IFC "The Value of Reputation" (2020)
• ILO Skills Recognition Framework (2023)
• World Bank "Mobile-Based Reputation Systems" (2021)
• Kenya NITA (National Industrial Training Authority) trade classifications

⚠️ DEMO — Production requires Huduma Number / NIIMS integration for identity verification.

---

© 2026 Gabriel Mahia / AI Kung Fu LLC · MIT License

Part of the East Africa Coordination Stack

This MCP server is one of 32 tools in 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 32 servers: pypi.org/user/gmahia Live demo: coord-cascade-demo

IP & Collaboration

MIT licensed. Feedback via GitHub Issues only — pull requests are not accepted. Demo data is labeled DEMO and is not suitable for operational decisions. Full policy: docs/architecture/IP_POLICY.md. Security reports: see SECURITY.md.

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.