Generate, explain, review and debug code.
CodeGPT helps with the parts of development that are mechanical or unfamiliar: writing a query against a schema you paste in, explaining dense code, drafting tests, and proposing causes for a failing case. Everything it produces is a draft to be reviewed and run against tests.
The CodeGPT interface runs here. While it is being connected, the guidance below covers how to use it, what it handles well and where it falls short — the same information you would want before running a first job.
Input
Two CREATE TABLE statements for customers and orders, engine stated as PostgreSQL, request for customers with no orders in the last 90 days.
Result
A LEFT JOIN with a date filter in the ON clause and a NULL check, plus a note that placing the date filter in WHERE would silently convert it to an inner join.
Input
A Node.js TypeError with the full trace, the failing function, and the observed versus expected behaviour.
Result
Three ranked hypotheses, each explaining how it would produce that exact message, with the check to run for each rather than a single blind fix.
Code you paste is processed to produce your result and is not retained as a permanent record. Never paste credentials, API keys, private keys or production data. Reduce examples to minimal reproductions where your policy requires it.
The full statement is in our privacy policy.