BPS_API_KEY=your_key npx dml-bps-mcp

git clone https://github.com/Digimetalab/dml-bps-mcp
cd dml-bps-mcp
npm install
npm run build
BPS_API_KEY=your_key npm start

https://dml-bps-mcp.digimetalab.workers.dev/mcp

MCP Server URL: https://dml-bps-mcp.digimetalab.workers.dev/mcp

{
  "mcpServers": {
    "dml-bps-mcp": {
      "type": "http",
      "url": "https://dml-bps-mcp.digimetalab.workers.dev/mcp",
      "headers": {
        "X-BPS-API-Key": "your_api_key_here"
      }
    }
  }
}

{
  "mcpServers": {
    "dml-bps-mcp": {
      "command": "npx",
      "args": ["-y", "dml-bps-mcp"],
      "env": {
        "BPS_API_KEY": "your_api_key_here"
      }
    }
  }
}

claude mcp add bps -- npx -y dml-bps-mcp

{
  "mcpServers": {
    "bps": {
      "command": "npx",
      "args": ["-y", "dml-bps-mcp"],
      "env": {
        "BPS_API_KEY": "${BPS_API_KEY}"
      }
    }
  }
}

{
  "mcpServers": {
    "dml-bps-mcp": {
      "command": "npx",
      "args": ["-y", "dml-bps-mcp"],
      "env": {
        "BPS_API_KEY": "your_api_key_here"
      }
    }
  }
}

User: "Berapa angka kemiskinan Indonesia 2023?"

AI menggunakan: find_data(query="penduduk miskin", region="Indonesia", year="2023")

Proses internal (otomatis):
0. Intent Detection: "single_value" → find_data
1. Resolve "Indonesia" → domain 0000
2. Normalize: "berapa angka kemiskinan" → "kemiskinan" (stopwords-iso)
3. Cari subjek relevan → "Kemiskinan dan Ketimpangan"
4. Cari variabel → "Jumlah Penduduk Miskin" (var_id: 183)
5. Resolve "2023" → period ID 123
6. Ambil data → 25,9 juta jiwa
7. Result hints: "💡 Gini rasio: get_dynamic_data(var="98")"

Jika find_data gagal, AI bisa:
- find_variable(keyword="miskin") → lihat variabel yang tersedia
- list_strategic_indicators() → data headline terbaru
- search(keyword="kemiskinan") → cari tabel/publikasi terkait

User: "Bandingkan kemiskinan Jawa Timur dan Jawa Barat"
AI menggunakan: compare_data(query="kemiskinan", regions="Jawa Timur, Jawa Barat")

User: "Tren pengangguran Indonesia 2019-2024"
AI menggunakan: get_trend(query="pengangguran", region="Indonesia", start_year="2019", end_year="2024")

User: "10 provinsi termiskin"
AI menggunakan: get_ranking(query="kemiskinan", top_n=10, order="highest")

User: "Statistik pemeluk agama di Kab Jombang"
AI menggunakan: find_data(query="pemeluk agama", region="Kab Jombang")
→ Intent: "table" → find_data dengan static table fallback
→ Otomatis ambil tabel statis "Jumlah Penduduk Menurut Agama"
→ Result hints: "💡 Breakdown detail: list_static_tables(keyword="agama")"

"Berapa jumlah penduduk miskin Indonesia tahun 2023?"
"Bandingkan angka kemiskinan Jawa Timur vs Jawa Barat 2020-2023"
"Tren pengangguran Indonesia dari 2019 sampai 2024"
"10 provinsi dengan kemiskinan tertinggi"
"Peringkat IPM seluruh provinsi 2023"
"Cari BRS terbaru tentang inflasi"
"Data ekspor kopi Indonesia tahun 2024"
"Cari publikasi tentang statistik telekomunikasi"
"Cari teks tentang akses internet di dalam publikasi BPS"
"Berapa IPM Jawa Timur?"
"Pertumbuhan ekonomi Indonesia triwulan terakhir"
"Statistik pemeluk agama di Kabupaten Klaten"
"Distribusi penduduk per kecamatan di Jakarta"

git clone https://github.com/Digimetalab/dml-bps-mcp
cd dml-bps-mcp
npm install

npm run build          # Compile TypeScript
npm run test:unit      # Run unit tests (105+ tests)
npm run lint           # ESLint check
npm run typecheck      # TypeScript type check

# Dengan environment variable
BPS_API_KEY=your_key npm start

# Atau buat file .env (lihat .env.example)
cp .env.example .env
# Edit .env, isi BPS_API_KEY
npm start

# Install dan jalankan inspector
npx @modelcontextprotocol/inspector

# Di inspector UI:
# 1. Transport: stdio
# 2. Command: node
# 3. Args: dist/index.js
# 4. Env: BPS_API_KEY=your_key

# Test initialize
echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | BPS_API_KEY=your_key node dist/index.js

# Test find_data
printf '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}\n{"jsonrpc":"2.0","method":"notifications/initialized"}\n{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"find_data","arguments":{"query":"inflasi","region":"Indonesia"}}}\n' | BPS_API_KEY=your_key node dist/index.js

# Jalankan worker secara lokal
npm run dev:worker

# Test di terminal lain
curl -X POST http://localhost:8787/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "X-BPS-API-Key: your_key" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

src/
├── auth/           # API key & OAuth2 providers
├── client/         # BPS WebAPI & AllStats HTTP clients
├── config/         # Configuration & defaults
├── prompts/        # MCP prompt templates
├── resources/      # MCP resources (domain lists)
├── services/       # Cache, domain resolver, data formatter
│   ├── intent-detector.ts   # Intent detection (comparison, trend, ranking, table)
│   ├── learning.ts          # Persistent learning store + stopwords-iso
│   ├── domain-resolver.ts   # Fuzzy domain matching
│   └── data-formatter.ts    # Format BPS data to markdown
├── tools/          # MCP tool definitions (39 tools)
│   ├── smart.tools.ts      # find_data, find_variable (AI shortcuts + intent detection)
│   ├── analysis.tools.ts   # compare_data, get_trend, get_ranking
│   ├── dynamic-data.tools.ts  # Core data tools
│   ├── search.tools.ts     # Search with AllStats fallback
│   ├── allstats.tools.ts   # AllStats search & deep search
│   └── ...                  # Domain, publication, trade, etc.
├── transport/      # stdio transport
├── utils/          # Logger and error handling
├── index.ts        # CLI entry point (stdio)
├── worker.ts       # Cloudflare Worker entry point (HTTP)
└── server.ts       # MCP server factory