Ask any SQL database in natural language. Agent explores schema, writes SQL, self-corrects.
MCP server for text2sql-framework. Plugs into Claude Desktop, Cursor, Goose, or any other MCP-compatible assistant and lets it ask a SQL database questions in natural language.
The agent explores the schema, writes SQL, executes it against the real DB, and self-corrects on errors — no RAG layer, no schema descriptions, no pre-computed embeddings.
Out of the box, text2sql-mcp supports SQLite + Anthropic:
pip install text2sql-mcp
# or
uvx text2sql-mcp
For other databases or LLM providers, install with the matching extra so the right driver gets installed:
| You want… | Install command |
|---|---|
| SQLite (default) | uvx text2sql-mcp |
| Postgres | uvx 'text2sql-mcp[postgres]' |
| MySQL | uvx 'text2sql-mcp[mysql]' |
| Snowflake | uvx 'text2sql-mcp[snowflake]' |
| BigQuery | uvx 'text2sql-mcp[bigquery]' |
| OpenAI models | add openai, e.g. uvx 'text2sql-mcp[postgres,openai]' |
Set environment variables in your MCP client config:
| Variable | Required | Description |
|---|---|---|
TEXT2SQL_DATABASE_URL | yes | SQLAlchemy URL, e.g. sqlite:///mydb.db, postgresql://user:pass@host/db |
ANTHROPIC_API_KEY or OPENAI_API_KEY | yes | LLM provider key |
TEXT2SQL_MODEL | no | LangChain model id (default: anthropic:claude-sonnet-4-6) |
TEXT2SQL_INSTRUCTIONS | no | Business rules / hints, e.g. "Revenue = net of refunds." |
TEXT2SQL_EXAMPLES | no | Path to a scenarios.md file for the agent's lookup_example tool |
{
"mcpServers": {
"text2sql": {
"command": "uvx",
"args": ["text2sql-mcp"],
"env": {
"TEXT2SQL_DATABASE_URL": "sqlite:///mydb.db",
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}
goose configure
# Add Extension → Command-line Extension
# Name: text2sql
# Command: uvx text2sql-mcp
# Env: TEXT2SQL_DATABASE_URL, ANTHROPIC_API_KEY
query(question, max_rows=100) — ask the database a natural-language question. Returns {sql, data, error, row_count, tool_calls_made}.Under the hood this is a thin wrapper around text2sql-framework, which uses LangChain Deep Agents to do iterative tool-calling against a single execute_sql tool. See the framework README for benchmarks (19/20 on Spider zero-shot across 80 tables) and architecture details.
MIT
Source-derived launch command. Check the maintainer’s required arguments and credentials before running:
uvx text2sql-mcpMerge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.
{
"mcpServers": {
"io-github-cpenniman12-text2sql-mcp": {
"command": "uvx",
"args": [
"text2sql-mcp"
]
}
}
}Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.
Claude Desktop setup referencetext2sql-mcppypitext2sql works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.
~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.~/.cursor/mcp.jsonRestart Cursor for changes to take effect..vscode/mcp.jsonReload VS Code window for changes to take effect.~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect..mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.