git clone https://github.com/MindsetConsulting/sflight-mcp.git
cd sflight-mcp
bash scripts/setup.sh   # installs deps, builds MCP server, creates .mcp.json
cds watch

git clone https://github.com/MindsetConsulting/sflight-mcp.git
cd sflight-mcp
powershell -ExecutionPolicy Bypass -File scripts\setup.ps1
cds watch

git clone https://github.com/MindsetConsulting/sflight-mcp.git
cd sflight-mcp
npm install
cd mcp-server && npm install && npm run build && cd ..
cds watch

npm install -g @anthropic-ai/claude-code

cd mcp-server
npm install
npm run build
cd ..

{
  "mcpServers": {
    "cap-tools": {
      "command": "node",
      "args": ["<FULL-PATH-TO>/sflights-mcp/mcp-server/build/index.js"]
    }
  }
}

claude

You: Which airlines have the most passengers?
You: Show me all flights from Frankfurt to New York
You: What's the average ticket price by airline?
You: How many bookings were made in 2025?

sequenceDiagram
    participant User as You (Terminal)
    participant CC as Claude Code
    participant MCP as MCP Server (Node.js)
    participant Tool as Tool Handler

    User->>CC: "Which airlines fly to London?"
    activate CC
    Note over CC: AI decides which tool to call

    CC->>MCP: {"method": "tools/call", "params": {"name": "cap_cql_query", "arguments": {"sql": "SELECT ..."}}}
    activate MCP
    Note over CC,MCP: JSON-RPC over stdin/stdout

    MCP->>Tool: Execute cap_cql_query handler
    activate Tool
    Tool->>Tool: Boot CDS + SQLite in-memory
    Tool->>Tool: Load 24 CSV seed files
    Tool->>Tool: Run SQL query
    Tool-->>MCP: {rows: [...], count: 5}
    deactivate Tool

    MCP-->>CC: {"result": {"content": [{"type": "text", "text": "## Query Results\n..."}]}}
    deactivate MCP

    Note over CC: AI formats the answer in plain English
    CC-->>User: "5 airlines fly to London: Lufthansa, British Airways, ..."
    deactivate CC

sequenceDiagram
    participant CC as Claude Code
    participant HM as hana-cli MCP Server
    participant HANA as SAP HANA Cloud

    CC->>HM: tools/call: "tables" (list all tables)
    activate HM
    HM->>HANA: SQL: SELECT * FROM SYS.TABLES
    HANA-->>HM: Table list
    HM-->>CC: Formatted table list
    deactivate HM

    CC->>HM: tools/call: "inspectTable" (show columns)
    activate HM
    HM->>HANA: SQL: SELECT * FROM SYS.TABLE_COLUMNS WHERE TABLE_NAME = '...'
    HANA-->>HM: Column definitions
    HM-->>CC: Column details with types and keys
    deactivate HM

    CC->>HM: tools/call: "querySimple" (run SQL)
    activate HM
    HM->>HANA: User's SQL query
    HANA-->>HM: Result set
    HM-->>CC: Query results as markdown
    deactivate HM

    Note over CC,HANA: hana-cli requires HANA Cloud credentials<br/>(default-env.json or CF service binding)

sequenceDiagram
    participant User as You
    participant CC as Claude Code
    participant CAP as CAP MCP Server<br/>(this project)
    participant CDS as CDS Compiler
    participant SQLite as In-Memory SQLite
    participant CSV as 24 CSV Files

    User->>CC: "What's the busiest route?"

    Note over CC: AI has 16 cap_ tools available<br/>+ hana-cli tools if configured

    CC->>CAP: cap_db_schema (understand table structure)
    activate CAP
    CAP->>CDS: cds.load('*') — compile CDS model
    CDS-->>CAP: Compiled model (CSN)
    CAP->>SQLite: Deploy model to in-memory DB
    CAP->>CSV: Load all 24 seed data files
    CSV-->>SQLite: 5,000 flights, 10K bookings, ...
    SQLite-->>CAP: Schema with column names and types
    CAP-->>CC: Schema details
    deactivate CAP

    CC->>CAP: cap_cql_query (run the actual query)
    activate CAP
    Note over CAP: Uses cached model (0ms)
    CAP->>SQLite: SELECT CITYFROM, CITYTO, COUNT(*)<br/>FROM flights_Connections<br/>GROUP BY CITYFROM, CITYTO<br/>ORDER BY COUNT(*) DESC
    SQLite-->>CAP: Top routes by flight count
    CAP-->>CC: Formatted results
    deactivate CAP

    CC-->>User: "The busiest route is Frankfurt → New York<br/>with 47 connections across 8 airlines"

    Note over CAP,SQLite: No HANA needed — works offline<br/>with SQLite + CSV seed data

cf login -a https://api.cf.<YOUR-REGION>.hana.ondemand.com
# Example regions: us10, eu10, ap21
# Select your org and space when prompted

npx cds build --production
mbt build

cf deploy mta_archives/sflights-mcp_1.0.0.mtar

cf apps
# Note the URL for sflights-mcp-approuter
# Open: https://<your-approuter-url>/odata/v4/flights/$metadata

# In BAS terminal
cd mcp-server && npm install && npm run build && cd ..
npm install -g @anthropic-ai/claude-code
claude

# 1. Clone and install
git clone https://github.com/MindsetConsulting/sflight-mcp.git
cd sflights-mcp
npm install

# 2. Build the MCP server
cd mcp-server
npm install
npm run build
cd ..

# 3. Create your MCP config (see "Setting Up AI-Powered Data Exploration" above)
# 4. Start exploring!
cds watch        # Start the OData service (optional)
claude           # Start asking questions

# If you've already deployed, download the credentials from CF:
cf env sflights-mcp-srv
# Look for the VCAP_SERVICES section — copy the hana credentials

{
  "VCAP_SERVICES": {
    "hana": [
      {
        "name": "sflights-mcp-db",
        "label": "hana",
        "plan": "hdi-shared",
        "credentials": {
          "host": "your-hana-host.hanacloud.ondemand.com",
          "port": "443",
          "user": "YOUR_HDI_USER",
          "password": "YOUR_HDI_PASSWORD",
          "schema": "YOUR_HDI_SCHEMA",
          "certificate": "-----BEGIN CERTIFICATE-----\n...\n-----END CERTIFICATE-----",
          "driver": "com.sap.db.jdbc.Driver",
          "url": "jdbc:sap://your-hana-host:443?encrypt=true&validateCertificate=true"
        }
      }
    ]
  }
}

cds watch --profile production

# Bind to your deployed HDI service
cds bind --to sflights-mcp-db:sflights-mcp-db

# Run with hybrid profile (HANA DB + local server)
cds watch --profile hybrid

sflights-mcp/
  db/
    schema.cds              # 26 entities + 4 views (CDS data model)
    data/                   # 24 CSV seed data files
    package.json            # HDI deployer config
  srv/
    flights-service.cds     # OData V4 service definition (26 entities + 4 views + 1 function)
    flights-service.js      # Custom handler: getFlightsOnDate()
  app/router/               # Managed approuter for BTP deployment
    xs-app.json             # Route config with XSUAA auth
  scripts/
    setup.sh                # One-shot setup (macOS/Linux/WSL2)
    setup.ps1               # One-shot setup (Windows PowerShell)
    start-mcp-sse.sh        # Launch MCP server as HTTP/SSE — macOS/Linux/WSL2 (Path A)
    start-mcp-sse.ps1       # Launch MCP server as HTTP/SSE — Windows PowerShell (Path A)
  test/
    mcp-test.sh             # MCP Inspector CLI tests for all 19 tools
    mcp-test-questions.sh   # 10 example SQL queries via MCP
    auth-test.sh            # OAuth2 token + BTP API call test
    odata-queries.http      # OData V4 query test suite (VS Code REST Client)
    mcp-questions.http      # 20 AI questions for Claude Code
  docs/                     # Integration guides
    claude-for-excel.md     # Claude for Excel integration (local + BTP)
    integration-paths.md    # External MCP server comparison (3 paths)
  mcp-server/               # AI-Powered MCP Server (19 tools)
    src/
      index.ts              # MCP Server class with stdio transport
      cap-tools.ts          # Tool definitions with schemas and handlers
      cds-executor.ts       # Hybrid execution engine (direct CDS binary + cache)
      output-formatter.ts   # Data to markdown table conversion
    scripts/
      query-runner.cjs      # CDS + SQLite query runner (boots in-memory DB)
  mta.yaml                  # MTA deployment descriptor (3 modules)
  xs-security.json          # XSUAA security config (admin + viewer roles)
  package.json              # Root project dependencies

                    BTP Cloud Foundry
    ┌─────────────────────────────────────────┐
    │                                         │
    │  ┌──────────────┐   ┌───────────────┐   │
    │  │  Approuter   │──→│  OData V4     │   │
    │  │  (128 MB)    │   │  Service      │   │
    │  │  XSUAA Auth  │   │  (256 MB)     │   │
    │  └──────────────┘   └───────┬───────┘   │
    │                             │           │
    │                     ┌───────▼───────┐   │
    │                     │  HANA Cloud   │   │
    │                     │  HDI Container│   │
    │                     │  26 tables    │   │
    │                     │  4 views      │   │
    │                     │  24 CSV loads │   │
    │                     └───────────────┘   │
    │                                         │
    └─────────────────────────────────────────┘