Code review that knows your rules.

What is it?

Revvy is a VS Code extension that brings structured, AI-powered code review directly into your editor. Instead of generic AI feedback, it enforces a set of rules you define in YAML — tailored to your team's standards, domain, or industry requirements (e.g. safety-critical embedded C/C++, Python, and more).

Why not just use /review or type a prompt?

• Many developers are not good at writing prompts.
• Every review requires crafting a new prompt tailored to a specific tech stack or language.
• AI can't deliver a thorough review without specific rules or best practices — the kind that are usually defined at the company or team level.
• Rules vary from one developer to another, which is why Revvy uses hardcoded, shared rules stored in a single YAML file that the whole team uses.
• A clean UI that shows you the exact line of code with an inline comment when an issue or warning is detected.

That's all?

Not quite. Revvy also:

• Generates integration tests (functional + regression)
• Provides a Copy as Markdown button to copy all review comments as a ready-to-use prompt for fixing issues
• (Coming soon) A Fix button for one-click remediation

It works with the AI backend you already have — GitHub Copilot (no extra key needed), OpenAI GPT-4o, or Anthropic Claude — with automatic fallback.

Features

• Rule-driven reviews — define exactly what to check via YAML profiles; enable/disable individual rules; live-reload on save
• Multi-profile support — switch between profiles per project or file type; auto-detect from file extension
• Multi-AI-backend — Copilot (default), OpenAI GPT-4o, Anthropic Claude, with automatic fallback chain
• Multi-repo PR/MR review — paste GitHub PR and GitLab MR URLs together; reviews all diffs in one AI pass with cross-repo analysis
• Ticket / requirement validation — paste a Jira ticket ID; AI checks the code satisfies acceptance criteria
• Inline results — syntax-highlighted comments with click-to-navigate; filter by Error / Warning / Info
• Export as Markdown — one-click agent-ready report to paste into Copilot Chat or any AI coding tool

Demo

https://github.com/user-attachments/assets/a218557e-2baf-4a97-aa86-1e86361e306b

Quick Start

From the VS Code Marketplace:

Search for Revvy in the VS Code Extensions panel and install it.

From source:

git clone https://github.com/YonK0/revvy.git
cd revvy
npm install
npm run compile
# Press F5 in VS Code to launch the Extension Development Host

Once installed, click the Revvy icon in the Activity Bar. Three example profiles are created automatically on first use (c-embedded, yocto, python).

Writing Rule Profiles

Create .yaml files in .vscode-reviewer/profiles/. Rules are reloaded automatically on save.

profile:
  id: my-team
  label: "My Team Rules"
  version: "1.0.0"
  file_patterns:
    - "**/*.ts"

  rules:
    - id: NO_CONSOLE_LOG
      title: "No console.log in production code"
      category: "coding-standards"
      severity: warning      # error | warning | info
      enabled: true
      description: |
        console.log statements should not be merged to main.
        Use a structured logger instead.

See the bundled profiles in .vscode-reviewer/profiles/ for full examples.

MCP Servers (Optional)

Enables Jira ticket fetching and multi-repo PR/MR review. Copy .vscode/mcp.json and fill in your credentials — tokens are prompted at runtime and never stored on disk.

Requirements

• VS Code 1.96.0+
• Git in PATH
• One of: GitHub Copilot subscription or free tier / OpenAI API key / Anthropic API key

Architecture

src/
├── extension.ts      # Entry point — commands, WebView provider, YAML watcher
├── panelProvider.ts  # WebView UI — welcome, loading, results screens
├── reviewer.ts       # Core engine — prompt construction, AI call, response parsing
├── aiBackend.ts      # AI abstraction — Copilot, OpenAI, Anthropic + fallback chain
└── ruleLoader.ts     # YAML loader — file scanning, parsing, live-reload watcher

Contributing

Contributions are welcome — new rule profiles, AI backends, bug fixes, UI improvements. See CONTRIBUTING.md.