FAQ
The exact review scope, isolation boundary, platform support, model role, and current access state.
Runrail reviews an existing AI-generated Python candidate branch against a clean base. It records changed files, bounded path and diff checks, secret-like and binary-change checks, syntax-only Python compilation, and a local decision packet.
Runrail materializes tracked candidate files in a separate local review directory without checking the branch out in the source repository. The source checkout stays unchanged, and Runrail never applies the candidate automatically.
No. The built-in verifier performs a Python syntax preflight only and does not import project code. Your project tests, semantic review, security review, and final merge decision remain your responsibility.
Runrail Desktop includes launchers for Windows, macOS, and Linux. The CLI also supports WSL2. Node.js 20+, Python 3.11+ with venv, and Git are required.
No. The default candidate recommendation is deterministic, offline, and model-independent. DeepSeek is an optional BYOK second opinion after explicit network consent; it receives minimized metadata and has no execution, approval, apply, or merge authority.
The browser workspace can compare related local audit files without an account or purchase. The installed runrail compare command is a licensed Pro-and-above feature when subscriptions are available.
New subscriptions are currently paused. Anyone can try the bundled browser-only recorded audit at /workspace/ without an account or purchase. Existing subscribers use the subscriber download route; no new checkout is offered while access is paused.
