Two tools, and the order matters: ask what sqlglot supports, then transpile in those terms. That is what keeps an assistant from confidently naming a dialect that does not exist and getting a plausible-looking result. Errors surface as errors — an unsupported dialect or a syntax problem is reported rather than silently smoothed over, which for query translation is the only acceptable behaviour.
An MCP tool built on the sqlglot library for translating SQL between dialects — MySQL to PostgreSQL, and every other pair sqlglot knows. It reports unsupported dialects and syntax errors rather than returning quietly broken SQL.
- The full list of SQL dialects sqlglot supports, so the conversion is asked for in terms it recognises — `Dialects`
- Query conversion from a source dialect to a target dialect — `Transpiler`
- Explicit error reporting when a dialect is unsupported or the input does not parse
No account and no key. Two paths: locally with Python 3.13 or newer and `uv sync --frozen` then `uv run server.py`, or build the Docker image and run it with `docker`, passing `-i`, `--rm` and `--init`. Client configuration is a plain command entry either way. The package is `sqltranspiler` (0.1.0 in pyproject).
One command — docker run -i --rm --init -e DOCKER_CONTAINER=true sql-transpiler