CodeCrab – Local-first desktop app for AI code reviews

Software engineer Edy released CodeCrab, a local-first native desktop app that reviews pull requests by orchestrating a user's local CLIs so no source code leaves the laptop. CodeCrab builds a per-repository review profile from the local codebase, maps AI observations onto changed diff lines, runs read-only investigations of reviewer comments, and verifies assisted fixes against the user's own test suite such as pytest or npm test. The app integrates with existing Claude Code, Cursor and Jira setups and uses the user's own gh CLI login instead of org-wide OAuth permissions.

CodeCrab is a native AI desktop app that learns your codebase. It combines your skills with specialized CodeCrab review skills, orchestrating your local CLIs to review PRs with zero code leaving your laptop. Everything CodeCrab offers to accelerate your review workflow. PULL REQUEST CODE REVIEW Review your teammates' PRs faster Open any Pull Request from your colleagues and let CodeCrab walk through the diff with you. It maps AI observations directly onto the changed lines, so you catch bugs, risky patterns and regressions fast — before you hit "Approve". Full codebase context, not just diffs — reviews against your entire local repo, types and test suite. 100% local-first — catches logic bugs and regressions without uploading a single line to the cloud. Live PR review: file tree, diff viewer and inline AI observations Live investigation: understand every observation and reach the right fix MY PRs — FEEDBACK, FIX & REVIEW Turn feedback on your PRs into the right solution When a teammate leaves an observation on your Pull Request, CodeCrab runs a deep investigation for you. It digs through the code around every comment, connects it with your project's context and helps you understand — as precisely as possible — what the observation really means and what the correct solution looks like. Deep, read-only investigation of every reviewer observation on the diff. Assisted fix on your local branch, verified against your own test suite. Still working locally? Before you create the Pull Request, let CodeCrab review your local changes. It detects errors early, investigates every finding as deeply as needed, and helps you apply the right fix — so the PR you finally open ships cleaner, higher-quality code. Read-only review of your in-progress local changes, before anyone sees the diff. Investigate findings fast and apply the fix while the context is still fresh. Live pre-push review: local changes analyzed before the PR is created Built around your setup: repositories, skills, models and language — all configurable LOCAL INTEGRATION & EXTENSIBILITY — PLUGS INTO YOUR WORKFLOW It learns your codebase — and amplifies everything you already have Connect any repository, and CodeCrab learns its rules and patterns—building a per-repo review profile that powers specialized review agents and custom skills. It integrates seamlessly with your local skills, Claude Code, and Jira. We don't replace your setup—we complement it and let you extend it without limits. Learns each repo's patterns into a review profile and deep skills — reviews tuned to your codebase. Your skills + CodeCrab agents — reuse your existing local skills, combine them with ours, and extend as far as needed. Adaptive Codebase Profiles: Learns each repository's unique rules and patterns to build a custom review profile tuned specifically to your project. Limitless Customization: Plugs into your existing Claude Code, Cursor, and local custom skills—allowing experienced engineers to enforce their own standards without replacing their workflow. CodeCrab vs. cloud review bots The difference isn't the model — it's where your source code ends up. Capability / Feature CodeCrab Local-First Cloud SaaS Bots Code Privacy & Security Never leaves your machine 100% Client-Side Uploaded & processed on vendor cloud servers GitHub Permissions Your own CLI login gh Requires org-wide OAuth admin permissions Execution Transparency Live Execution Console real-time stdout/stderr Opaque "black box" cloud pipeline Tool Orchestration Uses your local Claude Code & Cursor setups Locked to vendor's fixed model wrappers Local Verification Executes your local test suite pytest, npm test Cannot run or verify local test suites Performance & Speed Instant local app native execution Queued background cloud runners WHY I BUILT CodeCrab Crafted by engineers who review code daily. Hi, I'm Edy — software engineer with 14+ years of experience, including years building systems at Google and Pinterest. As software engineering moves faster and AI toolsets generate code at unprecedented speeds, our primary bottleneck has shifted from writing code to reviewing Pull Requests efficiently. I built CodeCrab initially as a personal tool to solve my own workflow headaches: I needed a way to perform deep, Staff-level code reviews fast, without uploading sensitive private code to third-party servers. By making it a 100% Local-First Desktop App, CodeCrab gives you total privacy and complete control over your review pipeline. It reuses the local AI subscriptions you already own like Claude Code or Cursor with zero server markups or hidden fees. CodeCrab is currently in Public Beta version. I am actively refining it based on real engineering workflows, and I'd love to hear your feedback as we build the ultimate client-side review experience together.