CodeVetter

Product history

What changed, and what it means.

A curated record of shipped CodeVetter outcomes. Roadmap work stays in GitHub Issues; this page records what users can rely on today.

  1. CodeVetter focused on executable verification

    • The product now centers on proving whether coding agents completed a task correctly with reproducible runtime evidence.
    • CLI, MCP, and machine-readable verification bundles are the primary surfaces; the desktop app remains a local viewer.
  2. The verification workbench reached a complete local loop

    • Reviews can retain local sessions, recover managed runs, record intent closure, and produce evidence-owned artifact previews.
    • A public agent-PR corpus and redacted performance receipts make regression and reliability comparisons reproducible.
  3. Deterministic Review and Agent PR X-Ray shipped

    • Broad reviews now cover every changed file through bounded, checkpointed verification units.
    • Completed reviews can export fail-closed JSON, Markdown, and self-contained offline HTML evidence.
  4. Work became conversation-first

    • Codex and Claude sessions gained provider-aware controls, clear attention states, searchable history, and safe archiving.
    • A persistent Plan, Build, Review, Verify, Done board keeps live sessions attached to repository context.