The Code Review Workbench.

LaReview is a dev-first code review workbench for complex changes. It turns a PR or diff into a structured review plan, grouped by flows and ordered by risk, so you can review with depth and understanding.

Unlike auto-review bots that post comment spam, LaReview is a local-first tool that works with your existing AI coding agent to help you ship high-signal feedback.

LaReview is designed for a local-first, secure, and focused review experience.

Screenshot	Feature
	AI-Powered Planning Input a PR (GitHub/GitLab) or diff. LaReview acts as a staff engineer, looking for flows (e.g. auth, api) and risks to build a structured review plan.
	Review Summary Get a bird's-eye view of your review with Issue Checklist, Feedback items, Review Tasks grouped by flow, and a Files Heatmap showing risk distribution.
	Structured Task Tree Navigate the review as a hierarchical tree. Mark tasks as To Do, In Progress, or Done. Attach contextual notes to specific tasks or lines of code to keep track of your thoughts.
	Task-Focused Diffs Review one logical concern at a time. The workbench isolates relevant code hunks for each task, eliminating noise.
	High-Signal Framework The AI proactively identifies bugs and authenticates them against your rules. No comment spam—just focused feedback threads anchored to specific lines.
	Local Context (Zero Data Leaks) Link local Git repos to give the agent full access to search your codebase. Works with your existing AI coding agent (Claude, etc) and keeps data on your machine.
	Visual Diagrams Visualize the flow. Automatically generate diagrams to see architectural changes before you read a single line of code. (Requires D2).
	Team Rules Define rules like "DB queries must have timeouts" or "API changes need a migration note" to enforce standards automatically.
	Learning Patterns The AI learns from rejected feedback. Mark suggestions as "ignored" during reviews, then analyze rejections to discover patterns that calibrate future reviews—fewer nitpicks, more signal.
	Git Host Sync Submit your review feedback directly to GitHub or GitLab PRs with automatic summary generation.
	Export & Share Export your review summary to Markdown with diagrams and code insights. Copy to clipboard or save to file.
	CLI Support Launch reviews from the terminal: lareview, lareview pr owner/repo#123, git diff | lareview, or lareview --agent claude.

1. Input
   • Run lareview from terminal (opens GUI with current repo linked), or
   • Paste a unified diff, or
   • Paste a code review link from GitHub or GitLab, or a PR URL
2. Fetch (for PRs)
   • LaReview fetches the change locally and prepares the review workspace. No intermediate servers.
3. Generate a review plan
   • AI generates a review plan grouped by flows and ordered by risk. You can customize the agent (Claude, OpenAI, etc).
4. Review
   • Review task by task with your AI coding agent.
   • Ship focused feedback that authors can act on, without comment spam.

LaReview includes a CLI that integrates naturally into your developer workflow:

# Open GUI with current repository linked
lareview

# Review changes between branches
lareview main feature

# Pipe a diff directly to the GUI
git diff HEAD | lareview

# Review a GitHub/GitLab PR
lareview pr owner/repo#123

# Specify an agent
lareview --agent claude

The CLI launches the GUI in the background and passes your diff/repo info seamlessly.

Claude	Codex	Gemini	Grok	Kimi	Mistral	OpenCode	Qwen

• Rust nightly (see rust-toolchain.toml)
• GitHub CLI (optional):
  • Install: brew install gh (or your OS package manager)
  • Auth: gh auth login
• GitLab CLI (optional):
  • Install: brew install glab (or your OS package manager)
  • Auth: glab auth login
• D2 (optional): For visual diagram rendering. Install: brew install d2 (or see d2lang.com)
• AI agent configured on your machine
  • Learn more: https://agentclientprotocol.com/overview/introduction

Install LaReview via Homebrew with a single command:

# First, tap the repository
brew tap puemos/tap
# Then install
brew install --cask lareview

brew install puemos/tap/lareview

1. Go to the Releases page and download the latest binary for your operating system.
2. Follow the instructions for your OS below.

1. Download the lareview-macos-*.zip asset and unzip it. You should see LaReview.app.
2. Drag LaReview.app into /Applications.
3. If macOS blocks it on first run, open System Settings → Privacy & Security and allow it.
4. To use from terminal, add to PATH (or use the CLI Installation button in Settings):

   echo 'export PATH="$PATH:/Applications/LaReview.app/Contents/MacOS"' >> ~/.zshrc

1. Install required dependencies:

   # Debian/Ubuntu
   sudo apt-get update
   sudo apt-get install -y \
     libwebkit2gtk-4.1-0 \
     libappindicator3-1 \
     librsvg2-2 \
     libxdo3
   
   # For other distros, install equivalent WebKit2GTK 4.1 packages

2. Download the lareview-linux.tar.gz file and extract it.

   tar -xzvf lareview-linux.tar.gz

3. Make the binary executable and move it to your path.

   chmod +x ./lareview
   sudo mv ./lareview /usr/local/bin/lareview

LaReview runs natively inside WSL2 as a Linux app. Follow the Linux installation steps above (including dependency installation).

Optional: For better Windows integration (opening URLs in your Windows browser), install the wslu package:

# Ubuntu/Debian
sudo apt install wslu

# Other distros: see https://wslutiliti.es/wslu/install.html

If wslu is not installed, LaReview will fall back to using cmd.exe for opening URLs.

Launch LaReview and go to Settings → CLI Installation for a one-click setup to add lareview to your PATH.

cargo run

scripts/build_macos_app.sh

# Debian/Ubuntu
sudo apt-get update
sudo apt-get install -y libxkbcommon-dev libxkbcommon-x11-dev

1. Start the app:

   lareview  # From terminal with repo linked
   # OR
   cargo run  # From source

2. Open GENERATE

   • Paste a unified diff, or a GitHub/GitLab PR like owner/repo#123 (or a PR URL)
   • Or use the terminal: lareview pr owner/repo#123

3. Pick an agent and click generate

4. Switch to REVIEW

   • Work through tasks, add notes, and track status

Input:

• owner/repo#123
• https://github.com/owner/repo/pull/123
• https://gitlab.com/owner/repo/-/merge_requests/123

Paste a unified diff directly:

diff --git a/src/lib.rs b/src/lib.rs
index 123..456 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -1,3 +1,8 @@
+pub fn example() {
+  // ...
+}

LaReview stores review state locally.

• macOS DB: ~/Library/Application Support/LaReview/db.sqlite

• Linux DB: $XDG_DATA_HOME/lareview/db.sqlite or ~/.local/share/lareview/db.sqlite

• Override DB path:

  LAREVIEW_DB_PATH=/path/to/db.sqlite <RUN_COMMAND>

Executable discovery (GUI apps can start with a minimal PATH):

• LaReview hydrates PATH from your login shell when launched outside a terminal (macOS/Linux). Restart after shell PATH changes.
• You can override individual agent executables in Settings → Agent Settings.

Wipe local state:

• Delete the DB file listed above.

• Toolchain: nightly Rust with rustfmt and clippy components (edition 2024; see rust-toolchain.toml)

• Run the app: cargo run

• Reset/seed sample data:

  If you need more advanced control, use the CLI.

  CLI Reference

• Checks:

  • cargo fmt -- --check
  • cargo clippy --all-targets --all-features -- -D warnings
  • cargo test

• Logging:

  • LaReview uses the log crate with env_logger.
  • Set RUST_LOG to control log levels:
    • RUST_LOG=debug cargo run - Full debug output
    • RUST_LOG=acp=debug cargo run - ACP agent debug output only
    • RUST_LOG=error cargo run - Errors only
  • Logs appear in the terminal when running from source.

• Tests:

  • Unit tests: Located alongside modules in src/ (usually as tests.rs)
  • Integration tests: Located in the root tests/ directory
  • Run all tests: cargo test

• Supply chain:

  • cargo deny check (used in scheduled CI)

• Docs index: docs/README.md
• Architecture overview: docs/ARCHITECTURE.md
• Development guide: docs/DEVELOPMENT.md

Please read CONTRIBUTING.md before submitting a PR.

See SECURITY.md for reporting vulnerabilities.

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE)
• MIT license (LICENSE-MIT)

At your option.