Does it work with my AI coding agent?

That's the point. Refactron verifies a diff whoever authored it: Cursor, Claude Code, Copilot, Codex, a codemod, or you. Via MCP your agent can call it before it lands (early access).

What does UNPROVEN mean?

Your tests pass, but the changed lines aren't actually exercised by any test, so “green” proves nothing. We say so explicitly and list the tests to add, instead of hand-waving.

Does my code leave my machine?

No. Refactron runs locally against your real test suite in an isolated shadow tree. Nothing is sent to an external service to make the verdict.

How is this different from just running my tests in CI?

Refactron scopes the run to the blast radius of the change, tells you when the change is uncovered (UNPROVEN), lands atomically only on green, and produces a signed, reproducible record of why it was safe.

What ships today vs. what's coming?

Live today: the SQLAlchemy migration + verification report. Building in the open: the MCP server and the CI gate. Join early access to be a design partner.

All releases

May 16, 2026 · v0.2.1

Large-File Fix & Transform Coverage

analyze survives large files, var conversion is scope-correct, and format_to_fstring speaks the full printf grammar.

Om Sherikar

Bug fixes

Large files. analyze no longer crashes on files over ~32 KB: parsing uses tree-sitter streaming input, and a single unparseable file is skipped instead of aborting the run.
var_to_const_let. Reference resolution is now scope-correct, so same-named vars in different functions no longer drop the whole file.

Improvements

format_to_fstring. Converts the full printf grammar: %d, %.2f, %x, %o, %e, %g, and width and precision specifiers, not just plain %s.