mcp-turso-cloud

What is mcp-turso-cloud?

mcp-turso-cloud is an MCP server that integrates Turso databases with LLMs. It provides a two-level authentication system for organization-level and database-level operations, enabling LLMs to manage and query Turso databases directly.

How to use mcp-turso-cloud?

Configure it in your MCP client using npx with environment variables: TURSO_API_TOKEN, TURSO_ORGANIZATION, and optionally TURSO_DEFAULT_DATABASE, TOKEN_EXPIRATION, and TOKEN_PERMISSION. Then use the provided MCP tools to interact with your Turso databases.

Key features of mcp-turso-cloud

• List, create, and delete databases in your Turso organization
• Generate authentication tokens for specific databases
• Execute read-only SQL queries (SELECT, PRAGMA) safely
• Execute destructive SQL queries (INSERT, UPDATE, DELETE, etc.)
• List tables and describe table schemas
• Perform vector similarity search using SQLite vector extensions

Use cases of mcp-turso-cloud

• An LLM lists all databases in a Turso organization to get an overview.
• An LLM creates a new database and then generates a read-only token for it.
• An LLM runs a SELECT query to retrieve customer data from a database.
• An LLM inserts new records into a table after manual approval.
• An LLM performs a vector search on embeddings stored in a Turso database.

FAQ from mcp-turso-cloud

How does the two-level authentication work?

The server uses a Turso Platform API token for organization-level operations (managing databases) and database-specific tokens for database-level operations. Database tokens are generated automatically from the organization token and cached.

What environment variables are required?

You must set TURSO_API_TOKEN (your Turso Platform API token) and TURSO_ORGANIZATION (your Turso organization name). Optional variables include TURSO_DEFAULT_DATABASE, TOKEN_EXPIRATION (default '7d'), and TOKEN_PERMISSION (default 'full-access').

What security separation exists between query types?

Read-only queries (SELECT, PRAGMA) use the execute_read_only_query tool, while destructive queries (INSERT, UPDATE, DELETE, etc.) use execute_query. This allows different approval levels—read-only can be auto-approved, while destructive operations require explicit review.

What tools are provided for database operations?

Database tools include list_tables, execute_read_only_query, execute_query, describe_table, and vector_search. Organization tools include list_databases, create_database, delete_database, and generate_database_token.

What are the dependencies or runtime requirements?

The server runs via Node.js and is distributed as an npm package. It uses the Model Context Protocol, Turso, and libSQL. You can run it with npx mcp-turso-cloud after setting environment variables.