MCP-Turso
# mcp-turso
A Model Context Protocol (MCP) server that provides access to the Turso-hosted LibSQL databases. Currently, the server provides the following functionality:
- Retrieving a list of tables in a database
- Retrieving a database's schema
- Retrieving the schema of a table
- Performing SELECT queries
## Configuration
### With Claude Desktop
Add this to your `claude_desktop_config.json`:
```json
{
"mcpServers": [
"turso": {
"command": "npx",
"args": ["-y", "mcp-turso"],
"env": {
"TURSO_DATABASE_URL": "your_url",
"TURSO_AUTH_TOKEN": "your_token"
}
}
]
}
```
You will need an existing database to continue. If you don't have one, [create one](https://docs.turso.tech/quickstart). To get the database URL via the Turso CLI, run:
```bash
turso db show --url <database-name>
```
Then get the database authentication token:
```bash
turso db tokens create <database-name>
```
Add those values to your configuration as shown above.
### With Cursor
To configure the Turso MCP server with Cursor, add the following to your Cursor settings:
1. Open Cursor and go to Settings (⚙️) > Settings (JSON)
2. Add the following configuration to your settings JSON:
```json
"mcpServers": {
"turso": {
"command": "npx",
"args": ["-y", "mcp-turso"],
"env": {
"TURSO_DATABASE_URL": "your_url",
"TURSO_AUTH_TOKEN": "your_token"
}
}
}
```
Replace `your_url` and `your_token` with your Turso database URL and authentication token as described in the previous section.
### Logging
The server includes a custom logger for debugging outside of Claude Desktop. By default, this logger writes to `<parent-dir>/logs/mcp-turso.log`, where `<parent-dir>` is the parent directory of directory containing the `mcp-turso` script. In other words, if the path to `mcp-turso` is `~/foo/bin/mcp-turso`, the logs will be at `~/foo/logs/mcp-turso.log`. If running with NPX as above, the default logs will be:
```
~/.npm/_npx/<npx-dir-name>/node_modules/mcp-turso/logs/mcp-turso.log
```
If you would like to specify a custom path, you can include a `--logs` flag with an **absolute posix path** in the server's configuration:
```json
{
"mcpServers": [
"turso": {
"command": "npx",
"args": ["-y", "mcp-turso", "--logs", "/Users/<username>/path/to/dir/mcp-logs.log"],
"env": {
"TURSO_DATABASE_URL": "your_url",
"TURSO_AUTH_TOKEN": "your_token"
}
}
]
}
```
The path to the log file (default or custom) is always logged to `stderr` when the server is created. For Claude desktop, this will show up in your server logs in `~/Library/Logs/Claude`.
_Note_: Right now, I haven't implemented specifying a custom logging file for Windows, but this is coming.
## Server Capabilities
The server provides the following tools:
- `list_tables`
- Get a list of all the tables in the database
- No input
- Returns: an array of table names
- `get_db_schema`
- Get the schemas of all tables in the database
- No input
- Returns: an array of SQL creation statements
- `describe_table`
- View schema information for a specific table
- Input:
- `table_name` (string): Name of table to describe
- Returns: Array of column definitions with names and types
- `query_database`
- Execute a SELECT query to read data from the database
- Input:
- `sql` (string): The SELECT SQL query to execute
- Returns: Query results as an object of type `{ columns: string[]; rows: Record<string, unknown>[]; rowCount: number; }`
## Todo
- [ ] Add the ability to specify a custom log file on windows
- [ ] Add more query tools
## License
MIT License - see the [LICENSE](LICENSE) file for details.
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose with no overlap: describe_table focuses on a single table's schema, get_db_schema covers all tables' schemas, list_tables simply enumerates table names, and query_database handles data retrieval. An agent can easily differentiate these based on their specific scopes.
All tools follow a consistent verb_noun pattern (describe_table, get_db_schema, list_tables, query_database) with clear, descriptive names. The naming is uniform and predictable, making it easy for agents to understand the action and target.
With 4 tools, this server is well-scoped for database interaction, covering essential operations like schema inspection, table listing, and data querying. Each tool earns its place without feeling excessive or insufficient for the domain.
The tools provide solid coverage for read-only database operations, including schema exploration and data querying. However, there are minor gaps for a full database management scope, such as lacking write operations (e.g., insert, update, delete) or administrative functions, which agents might need to work around.