Like it? Buy me a coffee.
Virtual Peer Review is free and open source. If it saves your team time, I'd appreciate the caffeine.
AI-powered code review that runs locally in your IDE. Catches security holes, logic errors, and race conditions before any human reviewer ever sees the code.
Finds the things automated tools miss. The kind of stuff that costs you hours in production.
Catches logic errors, off-by-one mistakes, wrong conditionals, and incorrect assumptions that slip through every unit test you wrote.
Identifies injection risks, auth gaps, privilege escalation paths, parameter tampering, and business logic abuse vectors.
Flags N+1 queries, unnecessary allocations in hot paths, and quadratic algorithms hiding in innocent-looking loops.
Point it at a folder of Markdown files with your org's standards. They become part of every review. Automatically enforced, zero nagging.
Block commits with unresolved findings. Bugs die in your terminal, not in production at 2am on a Saturday.
Runs on a local IPC socket. Your code never leaves your machine. No cloud, no telemetry, no "we train on your data" nonsense.
Every finding is categorised. Filter by what matters to you.
One service, every IDE. Same findings everywhere.
Install from your IDE's marketplace for automatic updates, or download manually.
Webview panel with findings list, click-to-navigate, and inline settings.
Native tool window with Kotlin/Swing UI, severity grouping, and notes.
Uses your existing CLI tools or a local llama.cpp server. The tool is free — bring whatever model you already pay for.
Requires a working local CLI (Claude, Codex, Kiro, OpenCode) or a running llama.cpp instance. No API keys entered — it delegates to whatever auth your CLI already has.
Anthropic's Claude CLI
CLIOpenAI Codex CLI
CLIAWS Kiro CLI
CLIOpenCode CLI
CLIAny local model
Local HTTPGrab the latest release for your platform.
Or install the terminal client via npm:
npm install -g virtualpeerreview
Seriously. Two minutes.
Grab the extension from your IDE's marketplace, or npm i -g virtualpeerreview for the terminal.
Claude CLI, Codex CLI, Kiro, OpenCode, or a local llama.cpp server. If you can run it in your terminal, it works.
The service starts automatically. Findings appear as you save. That's it. No config files, no YAML, no ceremony.
"I built this to help my team catch real issues faster. Every PR sat in the review queue while obvious bugs waited to be found. Now the obvious stuff is caught instantly, and human reviewers focus on architecture and design. I hope it helps your team too."
— Matt Ratcliffe
Virtual Peer Review is free and open source. If it saves your team time, I'd appreciate the caffeine.
This is open source. PRs welcome — whether it's a bug fix, a new provider, an IDE improvement, or a better prompt. Or fork it and make it yours. No gatekeeping here.