VibeRails is an enterprise-class code review harness for Claude Code and Codex. Run detailed analysis on entire codebases, lightning fast triage, then dispatch AI agents to fix what you approve.
Want to see the output? View a sample HTML report .
The Workflow
Add local git repositories. VibeRails scans files, counts tokens, and tracks changes across your entire codebase.
→AI agents analyze code file-by-file across 17 categories, building context as they progress through your codebase.
→Structured findings across 17 categories. Triage with full code context, approve, reject, or tag for batch processing.
→Dispatch AI agents to implement approved fixes in bulk, with human-in-the-loop when guidance is needed.
Not Just Another PR Checker
Lots of code review tools check pull requests or other incremental changes. VibeRails does a complete file-by-file review of existing large codebases, gathering insights, checking every file and returning structured detail on every issue found. Use Triage Mode for fast human approval, then Fix Mode to dispatch agents to address the issues you approve.
Intelligent Discovery
VibeRails groups files intelligently, then analyses them sequentially - hundreds or thousands of source files, checking for security vulnerabilities, performance issues, bugs, dead code, and more. As it reviews, it builds up context: dependency patterns, conventions, risk areas. Each discovery feeds into subsequent file reviews for sharper detection.
Triage Mode
A purpose-built full screen UI for reviewing issues at scale. Syntax-highlighted code, severity tags, tabbed detail panels, and a keyboard-driven workflow that lets you process dozens of findings without touching the mouse.
J/K navigation, single-key approve/reject/defer. Process dozens of findings without touching the mouse.
Overview, Impact, Fix, Code, and Details. Five perspectives on every finding, switchable with number keys.
Approve, reject, or defer at speed. Progress bar tracks your triage session in real time.
Annotate any finding with free-form notes. They persist with the issue and travel with exports.
Detection
Every issue classified, colour-coded, and prioritised. From critical security vulnerabilities to localisation gaps.
Critical
Code Health
Quality & Compliance
Batch Fix Sessions
Approve the issues that matter, then dispatch AI agents to fix them all at once. Real-time progress tracking, human-in-the-loop intervention when guidance is needed, and full git integration so you review every change before it ships.
Claude Opus 4.5 for fast broad discovery. GPT-5.2 Codex for deep verification. Different architectures catch different things.
AI agents pause when they need guidance. You provide direction, they continue. You stay in control of every change.
Use /viberails inside Claude Code to trigger reviews, check status, and triage issues without leaving your terminal.
Programmatic access to reviews, issues, and fix sessions. Integrate VibeRails into dashboards and custom workflows.
Adoption
Point VibeRails at a legacy codebase. In one session it produces a structured HTML report of every security vulnerability, performance issue, and bug found - across the entire codebase. Hand it to your CTO. Export to CSV for the compliance team. Let the results speak for themselves.
Your Keys. Your Models.
VibeRails orchestrates the AI CLIs you already have installed: Claude Code (Anthropic) and Codex (OpenAI). Use API keys for pay-as-you-go pricing, or subscription plans like Anthropic's Max Plan for unlimited usage. Your keys, your billing, your provider. VibeRails runs no servers between you and the AI.
Proof and Governance
Clear per-developer pricing, plus an enterprise launch pack for air-gapped rollouts. No hidden usage multipliers.
Data-flow boundaries, local/air-gapped deployment modes, and operational controls in one place for security review.
A practical implementation guide for introducing VibeRails into legacy engineering workflows, with pilot structure and governance checkpoints.
FAQ
VibeRails orchestrates your existing Claude Code and Codex CLI installations. When you run a review or fix session, relevant code is sent directly from your machine to the AI provider configured in those tools. VibeRails itself does not receive or store your source code.
Most tools review pull requests or incremental changes. VibeRails reviews your entire codebase file-by-file, accumulating architectural insights as it goes. It is designed for full-codebase audits, not line-by-line suggestions.
Yes. VibeRails orchestrates the CLIs you already have: Claude Code (Anthropic) and Codex (OpenAI). Your existing subscriptions, API keys and billing apply.
Any language that Claude and Codex can analyse: TypeScript, Python, Go, Rust, Java, C#, Ruby, and more. VibeRails is language-agnostic. The AI models handle the analysis.
VibeRails is capable of using a huge amount of compute (but the results are worth it!). We recommend subscription models like the Max Plan from Anthropic.
Right now VibeRails supports Opus 4.5 from Anthropic and GPT-5.2-codex from OpenAI (we'll add more frontier models as they get released).
Get product updates in your inbox.
No spam. Unsubscribe anytime.
Tell us about your team and rollout goals. We will reply with a concrete launch plan.