Sign up

Introduction

Jevlit reads the diff on a pull request and reports the bugs in it. It runs in CI, in your editor and from the terminal, and it reports through one model: a run that holds a list of findings.

What it does

A run takes the changed hunks of a pull request, not the whole repository, and asks a model what is wrong with them. What comes back is a finding per problem: a severity, a type, the file and the lines it sits on, and a report you can hand straight to a coding agent.

  • Diff-scoped — the unit of work is the change, so cost tracks the pull request rather than the codebase.
  • Agent-ready — every finding carries a prompt, so the fix starts where the report ends.
  • Reviewable — runs, findings and token spend all land in the console.

How a run works

  1. A pull request opens, or you run jevlit review locally.
  2. Jevlit collects the diff and the files it touches.
  3. Each hunk is reviewed; the findings are deduplicated against earlier runs on the same branch.
  4. The run appears in the run log, and a summary is posted back to the pull request.

Where to go next

If you want it running on a repository, start with the quickstart. If you want to know what it costs to run, the console tracks spend per service.