Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
Devin Review is an AI code review tool from Cognition that reads GitHub/GHES/GitLab PRs and posts structured, severity-ranked review comments. Its key differentiator is semantic file organization—grouping diffs by logical concern/intent rather than alphabetical order, with each hunk explained—which HN commenters have praised as a significant UX improvement over competitors.
Launched January 22, 2026 (free during early release; Team pricing now $500/seat/month, Enterprise quote-based), positioned alongside the full Devin AI agent. Built by Cognition (founded Nov 2023). As of May 27 2026 Cognition closed a $1B+ round at a $26B post-money valuation ($25B pre); ARR is ~$492M with enterprise usage up 50% month-over-month. Enterprise customers for the broader Devin platform include Goldman Sachs, Citi, Nubank, Dell, Cisco, Ramp, Palantir, Mercedes-Benz, Itaú, NASA, and the U.S. Army/Navy. Infosys strategic partnership announced January 2026.
Best fit: Teams wanting AI-augmented code review with minimal setup friction and enterprise-grade compliance inherited from Cognition's platform, plus a closed-loop auto-fix workflow. Less ideal for teams needing GitLab GA, Bitbucket/ADO support, or IDE-integrated review surfaces.
Adoption & Proof Points
- Parent company Cognition: closed $1B+ round May 27 2026 at $26B post-money ($25B pre), led by Lux Capital / General Catalyst / 8VC, with Founders Fund, Elad Gil, Ribbit, Atreides and others — a >2.5x step-up from the $10.2B post-money (Sept 2025 / $400M round). ARR ~$492M (13x YoY from ~$37M in May 2025); enterprise usage up 50% MoM for six months. 89% of Cognition's own code now written by Devin.
- Enterprise customers (Devin platform): Goldman Sachs (12,000-developer hybrid workforce pilot), Citi, Nubank, Itaú, Dell, Cisco, Ramp, Palantir, Santander, Mercedes-Benz, Mercado Libre, NASA, U.S. Army/Navy.
- Mercedes-Benz used Devin to compress an ~8-month COBOL modernization (200K+ LOC) into ~8 days; Itaú auto-fixes ~70% of security vulnerabilities with Devin.
- Windsurf acquisition (July 2025): ~$250M; SWE-1.5 model + Codemaps integration ongoing.
- Infosys strategic partnership (Jan 2026) for global enterprise deployment, starting with Financial Services.
- HN reception: Cautiously optimistic. Praised for semantic diff organization; criticized for brand baggage from the 2024 Devin demo controversy and the 'last 30%' problem (both apply to the full Devin agent, not Review specifically).
- Caveat: these are proof points for the Devin agent/platform, not Devin Review specifically — no Review-specific named production references yet.
Recommended Use Cases
- Augmenting human code review with automated severity-ranked bug detection on GitHub PRs
- Enforcing review standards across large teams via auto-triggered reviews on PR events
- Quick ad-hoc PR reviews using the URL swap method (replace github.com with devinreview.com)
- Closing the loop: auto-fixing flagged review comments and CI/lint failures until checks pass
- CI/CD integration via CLI (npx devin-review) or REST API v3 for automated review gates in pipelines
- Onboarding new reviewers by providing AI-generated, codebase-aware context on complex diffs
- Supply-chain / dependency-change scrutiny (demonstrated on the March 2026 malicious-axios attack)
Risks & Limitations
- No independent accuracy benchmarks for Review specifically (catch rate, false-positive/F1). DeepSource notes no AI-code-review vendor — Devin, CodeRabbit, Cursor BugBot — has published one; every self-published benchmark wins its own. Primary gate to Validated and the basis for the unvalidated-benchmarks cap.
- Not hands-on tested by us (handsOn=not_tested) — assessment rests on docs, vendor posts, and third-party reviews.
- GitLab not GA: interactive read-write PR review is preview only (since May 2026). No Bitbucket or Azure DevOps.
- GHES gaps: cannot post comments, submit reviews, or merge on GitHub Enterprise Server yet.
- No broad IDE-extension delivery for Review (web/dashboard/CLI/REST API only).
- Enterprise proof points are for the Devin agent/platform (Goldman, NASA, Itaú, etc.), not Devin Review specifically — no Review-specific named production references.
- Brand trust overhead: the early Devin agent faced credibility controversy over misleading 2024 demo videos, and the 'last 30%' reliability critique persists for the agent. Unrelated to Review but creates perception risk.
- Pricing: Team tier now public ($500/seat/mo); Enterprise remains quote-based. Verify total cost for high-volume PR review workloads.
- Company health is strong post the May 2026 $1B raise; the Aug 2025 ~30-staff layoff + ~200 buyouts were Windsurf-integration-related, not financial distress.
Capabilities & Integration
Bug Catcher: Categorizes issues by severity—red (probable bugs), yellow/orange (warnings), gray (FYI/commentary), plus copy/move detection. Performs read-only local execution (via CLI) for deeper analysis beyond surface-level diff scanning.
Access methods: Web UI via URL swap (replace github.com with devinreview.com in any PR URL), the app.devin.ai/review dashboard for managing reviews across repos, CLI via npx devin-review for terminal and CI/CD workflows, and a REST API v3 (including a GET /v3/enterprise/pr-reviews polling endpoint added May 2026) for programmatic triggering.
Closed-loop actions: The chat agent proposes code edits and applies them as commits to the PR branch. Devin can autofix incoming review comments (its own and other bots') and continues fixing lint/CI issues until all checks pass. Auto-merge toggle available from the merge bar.
Auto-review triggers: Configurable to run automatically on PR open, new commits, draft-to-ready transitions, or reviewer/assignee assignment; modes are Auto review / On PR creation / Manual. Respects project instruction files including REVIEW.md, AGENTS.md, CLAUDE.md, CONTRIBUTING.md, .cursorrules, .windsurfrules, .coderabbit.yaml, and custom glob patterns.
Semantic diff organization: Groups changed files by logical intent rather than alphabetical order and explains each hunk. Multiple HN commenters ranked this as the best AI code review UX, ahead of Claude and Codex reviews.
Platform coverage & limits: GitHub (full parity). GitHub Enterprise Server supported but cannot yet post comments, submit reviews, or merge. GitLab moved to interactive read-write PR review (May 2026) but remains preview, not GA. No Bitbucket or Azure DevOps. No broad IDE-extension delivery for Review.