DSE Analyst MCP — Análise Técnica e Cotações da Bolsa de Dhaka
Jakir123/dse-analyst-mcp
16 ferramentas de análise técnica + cotações DSE em tempo real via MCP
Ver no GitHubServidor MCP que conecta clientes de IA (Claude, Cursor, Windsurf) a dados em tempo real da Bolsa de Valores de Dhaka (DSE) e um motor completo de análise técnica com 16 ferramentas. Fornece cotações ao vivo, indicadores técnicos avançados (RSI, MACD, Ichimoku, Fibonacci) e scanner de mercado com estratégias de trading.
Instalação
npx convez-ai Jakir123/dse-analyst-mcp1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Análise técnica de ações da bolsa de Dhaka com indicadores avançados
- Monitoramento de cotações ao vivo e identificação de padrões de trading
- Scanning automático de mercado com estratégias momentum, swing e mean reversion
- Integração de dados financeiros em fluxos de trabalho com Claude, Cursor ou Windsurf
Veredito

Ferramenta robusta e bem estruturada para traders e analistas que operam na DSE, com cobertura completa de indicadores técnicos e integração MCP nativa. Limitação geográfica óbvia (apenas DSE), mas implementação técnica é sólida e documentação clara.
README
A tradução para o português deste README ainda está na fila.
DSE Analysis MCP Server
A custom Model Context Protocol server that connects any MCP-compatible AI client to live Dhaka Stock Exchange (DSE) data and a full technical analysis engine. Works with Claude Code, Cursor, Windsurf, Zed, and any other client that supports MCP. Built and tested primarily with Claude.
Project Structure
dse-analyst-mcp/
├── server.py ← MCP server — registers all 16 tools
├── dse_data.py ← DSE data layer (bdshare wrapper)
├── technical_analysis.py ← Indicators: RSI, MACD, BB, Ichimoku, Fib, etc.
├── requirements.txt ← Python dependencies
├── .mcp.json ← Claude Code MCP config (not committed — machine-specific paths)
├── .venv/ ← Python 3.11 virtual environment
├── CLAUDE.md ← Strategy reference (auto-loaded by Claude Code; adapt for other clients)
├── GUIDE.md ← User guide: queries, playbooks, indicator reference
└── README.md ← This file
Architecture
You (any MCP client — Claude Code, Cursor, Windsurf, etc.)
│ natural language question
▼
AI model ──── reads client MCP config ───────────────┐
│ │
│ picks tool │
▼ ▼
server.py (FastMCP) client config registers
16 registered tools server.py on startup
│
├── DSE Data tools → dse_data.py → bdshare → DSE website
└── TA tools → technical_analysis.py → pandas/numpy
Available Tools (16)
DSE Data Tools
| Tool | Description |
|---|---|
get_live_price(symbol) | Real-time quote — price, change, volume |
get_historical_data(symbol, days) | OHLCV candles for any date range |
get_market_summary() | DSEX, DS30, DSES index values |
get_top_gainers(n) | Top N gaining stocks today |
get_top_losers(n) | Top N losing stocks today |
get_all_live_prices() | Snapshot of all DSE-listed stocks |
get_company_info(symbol) | P/E ratio, EPS, NAV, market cap |
Technical Analysis Tools
| Tool | Indicators |
|---|---|
analyze_trend(symbol) | SMA 20/50/200, EMA 9/21/55, Golden/Death Cross |
analyze_momentum(symbol) | RSI (14), MACD (12,26,9), Stochastic, Williams %R, ROC |
analyze_volatility(symbol) | Bollinger Bands (20,2), ATR (14), Historical Volatility |
analyze_volume(symbol) | OBV, VWAP (20), Volume MA, volume ratio |
get_fibonacci_levels(symbol) | Retracements (23.6–78.6%) + extensions (127–261%) |
get_ichimoku_cloud(symbol) | Tenkan, Kijun, Senkou A/B, Chikou, cloud color |
get_pivot_points(symbol) | Classic pivot: Pivot, R1/R2/R3, S1/S2/S3 |
full_analysis(symbol) | All indicators combined + BUY/SELL verdict + confidence % |
Market Scanner
| Tool | Description |
|---|---|
scan_top_stocks(style, top_n) | Scans most active stocks, scores by strategy, returns ranked BUY list with entry/stop/target prices |
Styles: momentum swing long_term breakout mean_reversion all
Setup
First time (already done on this machine)
# 1. Create virtual environment with Python 3.11
python3.11 -m venv .venv
# 2. Install dependencies
.venv/bin/pip install mcp bdshare pandas numpy ta
# 3. Verify
.venv/bin/python3.11 -c "import server; print('OK')"
# 4. Configure your MCP client (see below), then restart it
Reinstall on a new machine
git clone <this repo>
cd dse-analyst-mcp
python3.11 -m venv .venv
.venv/bin/pip install -r requirements.txt
# Configure your MCP client and restart it
MCP config
The server uses stdio transport — your MCP client spawns server.py as a subprocess. Configure it with the absolute paths to your .venv and server.py.
Claude Code — create .mcp.json in the project root (not committed; gitignored):
{
"mcpServers": {
"dse-analysis": {
"command": "/path/to/.venv/bin/python3.11",
"args": ["/path/to/server.py"]
}
}
}
Cursor / Windsurf / Zed — add an equivalent entry in your client's MCP settings file. The command and args values are identical; only the config file location differs per client.
No internet connection is needed beyond what bdshare uses to scrape DSE.
Dependencies
| Package | Version | Purpose |
|---|---|---|
mcp | 1.27.2 | Anthropic MCP Python SDK |
bdshare | 1.2.1 | DSE live + historical data scraper |
pandas | 3.0.3 | DataFrame / OHLCV manipulation |
numpy | 2.4.6 | Numerical computing |
ta | 0.11.0 | Supplemental TA library |
Runtime: Python 3.11 (Homebrew). pandas-ta not used — all indicators are implemented natively in technical_analysis.py.
Troubleshooting
MCP server not appearing in your client
→ Claude Code: Cmd+Shift+P → "Developer: Reload Window"
→ Cursor / Windsurf: restart the application
→ Verify the command path in your client's MCP config points to the correct python3.11 inside .venv
Empty data / "No historical data found"
→ Try during DSE market hours: Sun–Thu, 10:00 AM – 2:30 PM BST (UTC+6)
→ Reduce the days parameter — some stocks have limited history
bdshare fetch errors
→ DSE's website structure occasionally changes; bdshare may need updating: .venv/bin/pip install --upgrade bdshare
ImportError on startup
→ Run .venv/bin/python3.11 server.py manually to see the full traceback