🚀 FastMCP Expense Tracker – Quick Start

▶️ Start MCP Inspector

Run the inspector with one command from the project root:

.\start-inspector.ps1

If the Inspector asks for connection details, use:

• Transport: streamable-http
• URL: http://127.0.0.1:8000/mcp

✅ The launcher disables proxy authentication for local development to avoid stale token issues.

⚠️ If port 6287 is already in use:

• Close any existing Inspector sessions
• Re-run the command

---

🖥️ Run the MCP Server

Start your server manually:

python main.py

🌐 Default Configuration

• Transport: HTTP
• Endpoint: http://127.0.0.1:8000/mcp

---

⚙️ Environment Configuration

You can override defaults using environment variables:

• MCP_TRANSPORT → (default: http)
• MCP_HOST → (default: 0.0.0.0)
• MCP_PORT → (default: 8000)

Example:

export MCP_HOST=127.0.0.1
export MCP_PORT=9000
python main.py

---

🗄️ Database Info

• Uses SQLite (async via aiosqlite)
• Database is stored in a temporary directory (safe + avoids permission issues)
• WAL mode enabled for better concurrency
• Auto-initializes on startup

⚠️ Note: Temporary storage means data may not persist permanently.

---

🧰 Available MCP Tools

➤ Expense Management

• add_expense → Add expense entry
• add_credit → Add income entry
• edit_expense → Update entry
• delete_expense → Remove entry

➤ Querying

• list_expenses → Fetch entries by date range
• summarize → Aggregate totals by category/type

---

📦 MCP Resource

• expense:///categories

  • Returns available categories
  • Falls back to default list if file is missing

---

🧪 Troubleshooting

❌ MCP not connecting

• Ensure server is running:

  python main.py

• Verify endpoint:

  http://127.0.0.1:8000/mcp
  

---

❌ Database errors

• Temp directory is used → ensures write access
• If issues persist, check system temp permissions

---

❌ Inspector issues

• Restart Inspector
• Close port 6287 conflicts
• Use correct transport: streamable-http

---

🚀 Next Steps

You can now:

• Connect this server to Claude / ChatGPT tools
• Build a frontend dashboard
• Deploy to Render / Railway
• Extend with analytics or AI features

---