mcp-turso

What is mcp-turso?

mcp-turso is a Model Context Protocol (MCP) server that integrates Turso databases with LLMs, using a two-level authentication system to manage organization-level and database-level operations directly from LLMs.

How to use mcp-turso?

Configure the server in your MCP client (e.g., Cline or Claude Desktop) by setting the required environment variables TURSO_API_TOKEN and TURSO_ORGANIZATION, and optionally TURSO_DEFAULT_DATABASE. Run with npx -y mcp-turso-cloud. Invoke tools like list_databases, create_database, execute_read_only_query, or vector_search via the LLM.

Key features of mcp-turso

• Two-level authentication (organization and database tokens)
• Organization operations: list, create, delete databases, generate tokens
• Database operations: list tables, read-only and destructive SQL queries
• Vector similarity search using SQLite vector extensions
• Schema description via describe_table tool
• Security separation between read-only and destructive queries

Use cases of mcp-turso

• Let an LLM list all databases in a Turso organization and manage them
• Query a specific database with read-only SQL (SELECT, PRAGMA) through an LLM
• Automate database schema exploration and table description in conversational workflows

FAQ from mcp-turso

What distinguishes read-only from destructive queries?

The server provides separate tools: execute_read_only_query for SELECT/PRAGMA (safe, can be auto-approved) and execute_query for INSERT/UPDATE/DELETE/CREATE (requires explicit approval).

What environment variables are required?

TURSO_API_TOKEN (Turso Platform API token) and TURSO_ORGANIZATION (organization name) are required. TURSO_DEFAULT_DATABASE, TOKEN_EXPIRATION (default '7d'), and TOKEN_PERMISSION (default 'full-access') are optional.

How are database tokens managed?

Database tokens are generated automatically using the organization token, cached for performance, and rotated as needed. Their expiration and permissions can be configured via environment variables.

What transports or authentication does the server use?

The server uses standard MCP tool-based transport (stdio via npx), authenticating via the Turso Platform API token for organization-level actions and database-specific tokens for database-level operations.

Are there any known limits or dependencies?

The server depends on the @libsql/client and Turso’s API. It runs via Node.js and requires network access to Turso. No other runtime dependencies are mentioned.