OpenReview
AI PR Review · #12 of 20 in category · #16 of 27 overall
Vercel Labs' self-hosted GitHub review bot: mention @openreview for Claude-powered inline suggestions; dormant beta since March 2026.
[ Where it fits ]
On the documented evidence, OpenReview suits teams that need the reviewer inside their own infrastructure.
[ Documented strengths ]
- Actionability. Inline GitHub suggestion blocks with one-click apply; commits simple fixes directly to the PR branch.
[ Documented gaps ]
- Dual-Workflow: Local vs. PR. GitHub App reviewing PRs only; no IDE, CLI, or pre-commit surface.
- Business Logic Validation. No issue tracker or MCP integration documented; reviews the PR in isolation.
- Continuous Learning. Emoji reactions gate applying fixes per PR; no persistent learning from feedback.
- Measurable ROI. No dashboards or metrics of any kind.
[ Facts ]
- Category
- AI PR Review
- Open source
- No — proprietary
- Pricing
- Free to run; you pay your own Anthropic API and Vercel hosting. Beta with no commits since March 2026 (~1.5K stars, not archived)
- Self-hosted
- Yes — full stack — Self-hosted only: deploy the Next.js app to Vercel with your own GitHub App credentials and Anthropic key.
- Platforms
- GitHub
- Model control
- BYOK Anthropic key on your own deployment; model hardcoded (Claude Sonnet 4.6), no provider choice without forking
- Last verified
- 2026-08-11
[ Against the 9 standards ]
Based on public documentation as of 2026-08-11. ✓ documented · ~ partial · ✗ not offered · ? undocumented. Undocumented scores zero — see the methodology.
Explores the codebase and runs project tooling during review; single repo, no persistent index.
Reviews only when mentioned, quiet by design; extensible skills in .agents/skills, no severity config.
No issue tracker or MCP integration documented; reviews the PR in isolation.
Emoji reactions gate applying fixes per PR; no persistent learning from feedback.
Runs project tooling (lint/format) as part of review; no documented test execution or preview environments.
Runs on your infra with your Anthropic key at zero markup; model hardcoded, no model independence.
Inline GitHub suggestion blocks with one-click apply; commits simple fixes directly to the PR branch.
[ OpenReview vs the alternatives ]
[ Closest alternatives ]
[ In our coverage ]
[ FAQ ]
Is OpenReview open source?
No. OpenReview is proprietary — you can use it, but you cannot read or fork the review engine.
Can OpenReview be self-hosted?
Yes — full stack. Self-hosted only: deploy the Next.js app to Vercel with your own GitHub App credentials and Anthropic key. Verified against the vendor's own documentation on 2026-08-11.
How much does OpenReview cost?
Free to run; you pay your own Anthropic API and Vercel hosting. Beta with no commits since March 2026 (~1.5K stars, not archived). Seat price is only part of the bill: OpenReview handles models as byok anthropic key on your own deployment; model hardcoded (claude sonnet 4.6), no provider choice without forking, which is what usually decides the real monthly cost.
How does OpenReview score against the 9-pillar AI code review standard?
3 out of 9. It fully documents 1 standard, partially documents 4, does not offer 4, and leaves 0 undocumented. The score is coverage of documented capability, not a measure of review quality.
What are the alternatives to OpenReview?
The closest tools in this directory are Gemini Code Assist, Bito, Cursor BugBot, PR-Agent. Each is scored against the same 9 standards, so the matrices are directly comparable.
Evaluating OpenReview?
Run it through the two-week trial protocol before you commit a team to it.