Give your reviews the context they deserve.

A review without context is just a linting pass. Drop a few configuration files in your repository and FriendlyReviewer will understand your architecture, ignore what doesn't matter, and write more relevant feedback.

📊 Review Reports: Helping Your Team Need Fewer Reviews

FriendlyReviewer doesn't just review your PRs, it also reviews its own work. Once enough reviews have accumulated, it generates an analysis report that helps you level up your practices so you need the AI a little less each time. And it's on the house.

📈 Quality trends & executive summary

See how your code quality evolves: average findings per review, recurring defect patterns, and the most critical issues to address.

📝 Recommendations for AGENTS.md

Concrete coding rules to add to your project guidelines, each one backed by real findings detected across your reviews, with the exact pattern, corrective rule, and expected impact.

🤖 Recommendations for .friendlyreviewer/helper.md

Adjustments for the AI reviewer itself: new checks to run, false positives to suppress, and edge cases to handle better. Copy them straight into your helper.md.

📏 Compliance

Existing project rules that were broken anyway, with the findings that show it. Deciding what to fix first stays with your team.

Reports are generated automatically once enough review data has accumulated (typically around 50 reviews), or after a month without one. They're sent by email to the account owner and available from their administration interface.

How It Works

FriendlyReviewer reads up to three optional files from the .friendlyreviewer/ directory at the root of your repository. Each one tunes a different aspect of the review:

ignore-reviewer.md

Which authors to skip entirely. List bot accounts (dependabot, renovate, etc.) so their automated commits don't trigger a review.

dependabot[bot]

ignore.md

Which additional files to skip. FriendlyReviewer already ignores common files (lock files, build artifacts, generated code) automatically. This file is for project-specific directories that only your repository knows about — legacy code, vendored dependencies, or any folder that should never be reviewed.

legacy/
vendor/

helper.md

What the AI should know. Injected into every review prompt. Contains a functional overview of your application, architectural rules, testing conventions, and project-specific best practices.

## Architecture
- Layer dependency flow: routes → tools → database

Bootstrap These Files In One Shot

helper.md is where the magic happens. Instead of writing it from scratch, paste this prompt into Claude Code, Cursor, or any AI coding assistant running from your repository:

I want to bootstrap FriendlyReviewer configuration for this project. Fetch the bootstrap guide: curl -s https://friendlyreviewer.fr/bootstrap-config.md Then follow its instructions to generate all 3 files in .friendlyreviewer/

The assistant will fetch the bootstrap guide, analyze your repository, and generate all three configuration files — including a helper.md with a functional overview of your app, architectural rules, and best practices extracted from your project's docs.

You can also download the bootstrap guide directly:

curl -s https://friendlyreviewer.fr/bootstrap-config.md