# # Prerequisites

#  Install dependencies
pip install -r requirements.txt

#  Install FastMCP globally (if not already installed)
pip install fastmcp

# # COMMAND CHEATSHEET

#  Run FastMCP directly for testing
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db fastmcp run sqlite_explorer.py

#  Test with inspector (if available)
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db fastmcp inspect sqlite_explorer.py

#  To install SQLite Explorer
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db fastmcp install sqlite_explorer.py - -name "SQLite Explorer"

#  To launch SQLite Explorer via a web-based testing interface. Run with `- -transport sse` for HTTP-based communication  
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db fastmcp dev sqlite_explorer.py

#  To set up the MCP server with Claude Desktop
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db fastmcp claude-desktop add sqlite_explorer.py - -name "SQLite Explorer"

#  Need to define the SQLITE_DB_PATH variable before running smithery playground 
SQLITE_DB_PATH=/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db smithery playground

After launching Smithery playground, we can now talk to the MCP server using this URL: https://smithery.ai/playground?mcp=https%3A%2F%2Fee09cd8f.ngrok.smithery.ai%2Fmcp

# # # For VSCode with Cline

#  Add this configuration to Cline MCP settings:
{
  "sqlite-explorer": {
    "command": "uv",
    "args": [
      "run",
      "- -with",
      "fastmcp",
      "- -with",
      "uvicorn",
      "fastmcp",
      "run",
      "/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/sqlite_explorer.py"
    ],
    "env": {
      "SQLITE_DB_PATH": "/Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db"
    }
  }
}

# # # Example output. MCP server provides four components. SQLite Explorer provides those tools.

Server
Name: SQLite Explorer
Generation: 2

Components
Tools: 3
Prompts: 0
Resources: 0
Templates: 0

Environment
FastMCP: 2.12.4
MCP: 1.15.0

This will open an interactive inspector where you can test the MCP tools:

• list_tables - to see what tables are in your database
• describe_table - to see the structure of a specific table
• read_query - to run SELECT queries on your data

# # Notes

Even though we're running the MCP locally, still have a web interface
For locally deployed MCP server SQLite Explorer, this is the MCP server URL that we can access as a client: http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=e01acfa4461c1af5227669c2c0051e7cf9c280fb61a43a146a672463441f4c1b# tools
In the future, once deployed, we'll have the same thing - HTTP or HTTPS, but instead of a local host, it'll be my server and the port. The rest will be something else.
Local host is the server name, so the server is a local host. But if deployed somewhere, then it'll be a different host name. The port serves as a channel to talk to multiple MCP servers. HTTPS is a secure HTTP.

# # # Sample Questions to Ask Your AI Assistant

Once the MCP server is connected, you can ask questions like:

• "What tables are in my database?"
• "Show me the structure of my financial data"
• "What are the first 10 rows from [table_name]?"
• "How many records are in each table?"
• "Show me all transactions over $1000"
• "What's the average amount in the transactions table?"

NOTE: The tables starting with SQLite are called system tables. We never created those tables.

# # Troubleshooting

#  Check FastMCP version and available commands
fastmcp - -help
fastmcp version

#  Verify your database file exists
ls -la /Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db

#  Test database directly with sqlite3
sqlite3 /Users/owner/claude-code/agentic-ai-learnings/hw3/sqlite-explorer-fastmcp-mcp-server/financial_data.db ".tables"

# # HW 3 Documentation

# # # Step 1

Cloned this GitHub repository: https://github.com/hannesrudolph/sqlite-explorer-fastmcp-mcp-server

# # # Step 2

Asked Claude Code: I was asked to evaluate the following deployment platforms: AWS, Vercel,
Railway, and Smithery for my MCP server. I want the simplest and cheapest
deployment. Which option should I select?

⏺ Based on your MCP server and requirements, I recommend Smithery.

Here's why:

Simplest: Smithery is specifically designed for MCP servers. It has the
most straightforward deployment process for this exact use case.

Cheapest: Smithery offers a free tier for MCP server hosting, making it
the most cost-effective option for your SQLite Explorer MCP server.

Other options comparison:

• AWS: Most complex setup, requires managing infrastructure, not
  cost-effective for simple MCP servers
• Vercel: Designed for web apps/APIs, not optimized for MCP servers,
  serverless model may not fit well
• Railway: Good platform but more expensive than Smithery's free tier,
  more general-purpose

For a read-only SQLite MCP server like yours, Smithery's MCP-specific
infrastructure is the best match for "simplest and cheapest."

# # # Step 3

Tested this locally via Cline. In VS Code, we configured the system settings to launch a web-based interface with the following JSON file. See URL: http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=a164e503687338cb23938baf05ae738ebe5cd0eaefa629e419cea7ef6ef51563# tools

# # # Step 4

smithery playground URL : https://smithery.ai/playground?mcp=https%3A%2F%2F143c4151.ngrok.smithery.ai%2Fmcp