Run Reviewate as part of your CI pipeline — no server or dashboard needed. Reviews run on every pull request and post comments directly.
GITHUB_TOKEN / GITLAB_TOKEN). The CLI fallback (gh / glab) is not available in container environments.Add this workflow to your repository:
# .github/workflows/ai-review.yml
name: AI Code Review
on:
pull_request:
types: [opened, synchronize]
jobs:
review:
runs-on: ubuntu-latest
container:
image: reviewate/code-reviewer:latest
steps:
- name: Run AI Review
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
run: |
reviewate ${{ github.repository }} \
-p ${{ github.event.pull_request.number }}
Required secrets: Set ANTHROPIC_API_KEY in your repository secrets. GITHUB_TOKEN is automatically provided by GitHub Actions.
Add this job to your .gitlab-ci.yml:
ai-review:
stage: test
image: reviewate/code-reviewer:latest
script:
- reviewate $CI_PROJECT_PATH
-p $CI_MERGE_REQUEST_IID
--platform gitlab
variables:
GITLAB_TOKEN: $GITLAB_TOKEN
ANTHROPIC_API_KEY: $ANTHROPIC_API_KEY
only:
- merge_requests
Required CI/CD variables: Set GITLAB_TOKEN (with api scope) and ANTHROPIC_API_KEY in Settings > CI/CD > Variables. See GitLab integration for token setup.
Use subcommands and flags to control behavior:
reviewate owner/repo -p 123 # review (default)
reviewate summary owner/repo -p 123 # summary only
reviewate full owner/repo -p 123 # review + summary
| Flag | Description |
|---|---|
-p / --pr | Pull request / merge request number |
--platform | github or gitlab |
--dry-run | Run without posting comments to the platform |
--debug | Enable debug mode with tracebacks and reasoning output |
--json | Output results as JSON |
Set the model via environment variables:
| Variable | Description |
|---|---|
ANTHROPIC_API_KEY | API key for Claude models (or use CLAUDE_CODE_OAUTH_TOKEN) |
REVIEWATE_REVIEW_MODEL | Review tier model (default: sonnet) |
REVIEWATE_UTILITY_MODEL | Utility tier model (default: haiku) |
REVIEWATE_BASE_URL | Custom base URL (e.g., LiteLLM proxy) |
See Model Configuration for the two-tier model system and all configuration options.