sql-steward runs on your own machine — the client starts it, so there is no endpoint to ping. 75 installs a week from pypi. Last commit 29 Jul 2026.
The agent never writes SQL: queries compile from a semantic layer you control, PII refused first.
The author archived this repository on GitHub, meaning it is no longer maintained.
Every tool that appeared, vanished or quietly changed what it asks for. Recorded since 20 August 2026. No other catalogue keeps this.
Quiet is not dead — but it is worth knowing when it wakes up, or when someone else takes it over. We watch the repository and tell you either way.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add sql-steward -e 'SQL_STEWARD_DB_URL=<SQL_STEWARD_DB_URL>' -- uvx sql-steward
{
"mcpServers": {
"sql-steward": {
"command": "uvx",
"args": [
"sql-steward"
],
"env": {
"SQL_STEWARD_DB_URL": "<SQL_STEWARD_DB_URL>"
}
}
}
}
[mcp_servers.sql-steward]
command = "uvx"
args = ["sql-steward"]
env = { SQL_STEWARD_DB_URL = "<SQL_STEWARD_DB_URL>" }
{
"mcpServers": {
"sql-steward": {
"command": "uvx",
"args": [
"sql-steward"
],
"env": {
"SQL_STEWARD_DB_URL": "<SQL_STEWARD_DB_URL>"
}
}
}
}
{
"servers": {
"sql-steward": {
"type": "stdio",
"command": "uvx",
"args": [
"sql-steward"
],
"env": {
"SQL_STEWARD_DB_URL": "<SQL_STEWARD_DB_URL>"
}
}
}
}
Put your own value in place of
<SQL_STEWARD_DB_URL> (SQLAlchemy URL of the database to read (read-only account recommended)).
The author’s registry entry marks it as required.
Optional settings from the registry entry, left out of the block above:
SQL_STEWARD_LAYER (Path to the semantic layer YAML (default: semantic.yaml)).
API-first semantic layer MCP server compiling YAML models into dialect-specific SQL
Every figure re-checked against SQL-computed facts; unsupported questions refused with the reason.
MCP server for natural-language queries against SAP CAP db-layer entities via real CQN/SQL JOINs.
An Oracle MCP server built from the PL/SQL, tables and statements you select. No run-sql tool
Read-only PostgreSQL over MCP. Writes refused at the parsed SQL, plus a READ ONLY transaction.
Local code search for AI coding agents: hybrid keyword + semantic + SQL over an index in your repo.
MCP server for one MySQL database: schema, read-only queries and guarded writes.
Read-only SQL on CSV files for AI agents. No writes, no network; --root confines file access.
Answers built from our own checks of this server.