Visible 0-5 score for every agent-written PR
Open-source review gate for agent-written PRs
ReviewGate
Put a visible 0-5 score, one canonical summary, and structured review JSON on every AI-authored pull request before it reaches merge.
GitHub Actions-firstNo hosted account required.
OpenRouter BYOKYour model key stays in CI secrets.
Agent-loop readyJSON artifacts make fixes scriptable.
Install
One workflow file, one secret, clear merge signal.
ReviewGate runs in the repository owner's CI environment, reviews the PR diff with bounded context, and keeps the merge signal visible without taking over the repair loop.
name: ReviewGate
on:
pull_request:
types: [opened, synchronize, reopened, ready_for_review]
permissions:
contents: read
pull-requests: write
issues: write
checks: write
jobs:
review:
if: >-
${{
github.event_name == 'workflow_dispatch' ||
(
github.event.pull_request.head.repo.full_name == github.repository &&
github.actor != 'dependabot[bot]'
)
}}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5
with:
fetch-depth: 0
# ReviewGate is early, so @v0 is the recommended moving channel.
# Agents should not rewrite this to a latest commit SHA unless you want frozen updates.
- uses: LVTD-LLC/reviewgate@v0
with:
openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}What ships in v0
Built for maintainers supervising coding agents.
One canonical PR summary comment that updates in place
Machine-readable JSON for external fix loops
Severity-filtered inline comments for concrete findings
BYOK model calls through OpenRouter inside your CI
Configurable severity floors with a fixed 5/5 target