🇨🇮 BRVM MCP Server

The first MCP server for West Africa's regional stock exchange.

Connect any AI agent — Claude, ChatGPT, Cursor, Copilot, LangGraph — to live market data from the BRVM (Bourse Régionale des Valeurs Mobilières), serving 8 UEMOA member states across West Africa.

Why?

The BRVM lists ~46 companies across Bénin, Burkina Faso, Côte d'Ivoire, Guinée-Bissau, Mali, Niger, Sénégal, and Togo — with a combined market cap of ~$27.8B USD. Yet it has no public API. This server bridges that gap, making BRVM data accessible to the AI ecosystem via the Model Context Protocol.

What's coming on the BRVM

• December 18, 2026: First derivatives market in francophone West Africa (futures on indices and stocks)

• 2026–2027: ETF listings

• 2026: Agricultural commodities exchange (BMPA CI) — cashew, maize, cola nut

This server is built to grow alongside these developments.

Features

Tools (7)

Resources

Prompts

Quick Start

Prerequisites

• Python 3.11+

• uv (recommended) or pip

Install & Run

# Clone
git clone https://github.com/dilanefk/brvm-mcp-server.git
cd brvm-mcp-server

# Install
uv sync

# Run (stdio — for Claude Desktop / Cursor)
uv run brvm-mcp

# Run (HTTP — for remote agents / web apps)
uv run brvm-mcp --transport streamable-http --port 8000

Connect to Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "brvm": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/brvm-mcp-server", "brvm-mcp"]
    }
  }
}

Restart Claude Desktop. You'll see the BRVM tools in the 🔨 menu.

Test with MCP Inspector

npx @modelcontextprotocol/inspector uv run brvm-mcp

Docker (remote deployment)

docker build -t brvm-mcp .
docker run -p 8000:8000 brvm-mcp

Example Conversations

Once connected, you can ask Claude:

"What's the BRVM market doing today?" → Calls get_market_summary(), returns full session data

"Show me Sonatel's stock price" → Calls get_stock_price("SNTS")

"Which BRVM stocks are from Sénégal?" → Calls search_stocks(country="Sénégal")

"Give me a full market report" → Uses the market_report prompt to orchestrate multiple tool calls

"Which stocks dropped the most today?" → Calls get_top_movers(), focuses on losers

Architecture

┌──────────────────┐     MCP (stdio/HTTP)     ┌──────────────────┐
│   AI Agent       │◄────────────────────────►│  BRVM MCP Server │
│ (Claude, GPT,    │                          │                  │
│  Cursor, etc.)   │                          │  FastMCP + Tools │
└──────────────────┘                          │  + Cache Layer   │
                                              └────────┬─────────┘
                                                       │ httpx
                                              ┌────────▼─────────┐
                                              │  Data Sources    │
                                              │  • afx.kwayisi   │
                                              │  • brvm.org      │
                                              │  • Rich Bourse   │
                                              └──────────────────┘

• FastMCP: Decorator-based MCP framework (official Python SDK)

• httpx: Async HTTP client for scraping

• BeautifulSoup4 + lxml: HTML parsing

• diskcache: TTL-based disk caching (5min default, respectful of sources)

• Pydantic: Structured data models

Roadmap

• v0.2 — Historical price data (OHLCV bars)

• v0.3 — BRVM official announcements / corporate actions feed

• v0.4 — BMPA CI commodities data (when available)

• v0.5 — Derivatives data (futures, options — post Dec 2026 launch)

• v1.0 — OAuth + rate limiting for multi-user deployment

• Register on MCP Registry

Data Sources & Fair Use

This server scrapes publicly available data. It implements:

• Caching (5min TTL) to minimize requests

• Polite User-Agent header identifying the project

• No authentication bypass — only public data

For real-time FIX feeds, contact BRVM directly: brvm.org/real-time-data-feed

Contributing

PRs welcome! Priority areas:

1. Additional data sources (eodhd.com, Rich Bourse)

2. Historical data support

3. Tests (pytest + VCR cassettes for scraper tests)

4. TypeScript port for Node.js MCP hosts

License

MIT

Built with ☕ in Mauritius by Dilane Fogué Kamga — Senior Software & AI Engineer, Financial Engineering MSc (WorldQuant University).

Bridging AI and African capital markets.