mcp -sqlite

This MCP server provides a SQLite-backed Model Context Protocol (MCP) instance that enables you to run SQL queries, manage schema, and generate business insights through an auto-updating memo at memo://insights. The server exposes six core tools: read_query to perform SELECTs and return results, write_query for INSERT/UPDATE/DELETE with a response that indicates affected rows, create_table to define new tables, list_tables to enumerate existing tables, describe-table to view a table's schema, and append_insight to add new business insights to the memo resource. Use the memo to collect and review insights discovered during data analysis, and leverage the mcp-demo prompt to guide interactive analysis of a specified business topic. Through the provided CLI you can run the server, feed it a database path, and then issue structured tool calls to read, modify, and analyze data while automatically updating insights in memo://insights.

Prerequisites:

• Docker or a compatible container runtime (for the Docker method) or a local environment with UV/uvx installed (for the UV/uvx methods).
• Access to a SQLite database or the ability to create one for testing (db_path).
• Basic familiarity with SQL and MCP tools.

Installation steps (choose one method):

1. UV/uvx (Python-based) method:

• Ensure uv/uvx is installed and accessible from your PATH.
• Place the mcp-server-sqlite entry point in a known location or reference it via your config.
• Provide a db-path when starting the server (for example, --db-path ~/test.db).

2. Docker method:

• Ensure Docker is installed and running.
• Pull or build the mcp/sqlite image as shown in the repository, e.g. docker build -t mcp/sqlite .
• Run the container with a bound path for the database using the provided command examples.

3. VS Code / Claude Desktop integration (optional):

• Copy the server configuration blocks into your claude_desktop_config.json or VS Code mcp.json as shown in the README examples.
• Provide the db-path input when prompted or via the file-based configuration.

Note: Replace placeholders like ${workspaceFolder}, ~/test.db, or /mcp/test.db with your actual paths.

Tips and caveats:

• Ensure the SQLite database file exists at the path you provide, or the server may attempt to create it depending on the interface used.
• The memo resource memo://insights is auto-updated whenever append_insight is invoked; design your prompts and analysis to trigger this tool as needed.
• If you use the Docker method, ensure the volume mappings align with where your database file resides and that file permissions allow the container to read/write.
• For quantitative work, frequently list_tables and describe-table to verify schema changes before running read_query/write_query operations.
• The mcp-demo prompt can help users explore the database domain by generating schemas and sample data tailored to a business topic.