Juror CLI
GitHub Action and CLI: several models review a pull request in parallel and duplicates merge into one report with a cost.
- 101 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Repository code and diffs are sent to model providers, and keys sit in GitHub secrets.
- The Action posts comments on pull requests under the repository token.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/jurorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Repository code and diffs are sent to model providers, and keys sit in GitHub secrets.
- The Action posts comments on pull requests under the repository token.
This is third-party code. Review the repository files before installing.
What it does
Juror runs in GitHub Actions on your runners and sends the diff to several models at once. The models read the repository with their own search tools, so no index or separate service is needed. Duplicate findings about one defect are merged by code similarity and a conservative referee, and the result is posted to the pull request as one comment with a merge score and a usage bill. You can start with `npx juror-ai init --set-secrets`, which writes a workflow with SHA-pinned actions and optionally uploads provider keys through `gh`. Locally a review runs with `npx juror-ai review --pr 1234 --repo owner/name`. There is also an optional post-merge browser check through Playwright and a remote MCP server for viewing findings.
Who it is for. Teams that want independent multi-model pull request review without a separate SaaS and with a visible price for every run.
Good fit when
- A single model reviewer misses defects and a second look from another model family is wanted.
- The team wants to see what reviewing a specific pull request cost.
- A review runs locally by pull request number without creating a workflow.
- After a merge the affected journeys are checked in a browser on a staging site.
Not a fit when
- An auto-fixer is needed; Juror only writes findings and does not edit code.
- A linter replacement or a code discussion chat is wanted.
- The repository is private and code cannot be sent to model providers.
Example request
Review pull request 1234 in owner/name with `npx juror-ai review`, show findings by severity and the run cost, but post nothing.Limitations
At least one provider key is required (`JUROR_OPENAI_API_KEY`, `JUROR_FIREWORKS_API_KEY` or `JUROR_OPENROUTER_API_KEY`), billed under the provider plan. Code and the diff go to the model provider. Provider secrets are not exposed to code from forks. The Greptile comparison in the README rests on one adjudicated pull request and is not a statistic. Connect browser checks only to a synthetic account without production data. Do not run the hosted service next to the Action for the same repository. The license is MIT.
How to disable. Delete `.github/workflows/juror.yml` and the `JUROR_*` secrets in the repository settings. For the CLI, simply stop running `npx juror-ai`.
FAQ
How do I add Juror to a repository?
Create `.github/workflows/juror.yml` following the README example or run `npx juror-ai init --set-secrets`. Then add at least one provider key to the repository secrets.
How many keys are needed?
One starts a review. A second provider turns the `fast` preset into a cross-model jury. The `starter` preset works with a single OpenRouter key.
Can findings be viewed from ChatGPT or Codex?
Yes, Juror Cloud has a remote MCP at `https://app.juror.dev/mcp` with OAuth sign-in. It shows findings and runs and starts a review after confirmation.
How does consensus mode differ from the default?
By default every non-duplicate finding is shown for recall. Setting `review.publish_mode: consensus` keeps only findings that several models agree on.
What is the post-merge browser check?
An optional mode: Juror plans the affected journeys and runs them on a staging site through Playwright. It is enabled with `npx juror-ai init --qa` and a staging URL.
Discussion
No questions or reviews yet.
Related
CLI toolsreviewd reviews GitHub, GitLab and BitBucket pull requests through Claude Code, Gemini or Codex CLI and comments on them
Jev-Code-Reviewer ranks changes in an agent PR as P0, P1 or P2 and retells them in plain language on the GitHub page.
A code-review CLI and plugin for agents: reads a git diff and writes line-level comments with context
A CLI that reviews code changes, fixes confirmed findings and repeats review before deterministic repository verification.
Two skills: a pull request review where two models argue, and a helper that works through review comments round by round
Robin reviews every pull request in your repository as a GitHub Action with your own AI key, including OpenRouter's free tier.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Juror (this card) | no | 101 | |
| Greptile skills | yes | 435 | skills that work through Greptile comments and fix them; Juror reviews with several models itself and does not edit code |
| code-review-skill | no | 2,256 | a code review skill for a single agent; Juror merges findings from several models into a report with a cost |
README in short
The package combines an Action, a CLI and a hosted service. There is no code index or storage, and every run gets a bill. It does not serve as an auto-fixer, linter or chat.