Feature Request: Graceful handling of character encoding errors in SQLite MCP server #124

Is your feature request related to a problem? Please describe.

When using Japanese text with SQLite database in the Claude Desktop Application, the MCP server crashes if character encoding is not properly configured. Specifically:

• Attempting to INSERT Japanese text causes the MCP server to crash
• Even for successful manual inserts ('apple', 'りんご') via direct SQLite operations, querying through the application shows corrupt characters:

  {'id': 1, 'word': 'apple', 'meaning_ja': '���'}
  

Describe the solution you'd like

The MCP server should handle character encoding errors gracefully without crashing. When encountering encoding issues, it should:

1. Continue running instead of crashing
2. Possibly return an error message indicating the encoding problem
3. Allow users to fix the encoding settings without needing to restart the server

Describe alternatives you've considered

Currently, the issue can be worked around by setting the PYTHONIOENCODING environment variable in the MCP server configuration:

{
  "mcpServers": {
    "sqlite": {
      "command": "uvx",
      "args": [
        "mcp-server-sqlite",
        "--db-path",
        "C:\\Users\\USER\\test.db"
      ],
      "env": {
        "PYTHONIOENCODING": "utf-8" 
      }
    }
  }
}

(Solution referenced from modelcontextprotocol/servers#65)

Additional context

Environment:

• Client Application: Claude Desktop Application
• Operating System: Windows 11
• Database: SQLite 3
• Database Encoding: UTF-8
• Content: Japanese text (UTF-8 encoded)

Technical Details:

SQLite itself is encoding-agnostic and treats data as byte strings. The character encoding interpretation happens at the interface level - in this case, Python's SQLite driver in the MCP server. Without proper encoding settings, Python may default to an incorrect encoding when reading from or writing to the database.