CodeCrab

Product Introduction

1. Definition: CodeCrab is a native desktop application (client-side AI code review tool) that performs automated, local-first analysis of Git diffs and Pull Requests. It functions as an orchestration layer for existing local AI command-line interfaces (CLIs) like Claude Code, Cursor, and GitHub Copilot CLI.
2. Core Value Proposition: It exists to provide engineers with fast, private, and context-aware AI-powered code reviews without uploading source code to third-party cloud servers. Its primary value is enabling secure, local AI code review that integrates into existing developer workflows while maintaining absolute code privacy and engineering control.

Main Features

1. Local-First Pull Request Review: CodeCrab clones the target repository locally and uses configured AI CLIs to analyze the diff against the full codebase context, including types and tests. Observations are mapped directly onto changed lines in a native diff viewer. The process is 100% client-side; code never leaves the developer's machine.
2. Deep-Dive Investigation & Assisted Fix: For each AI-generated observation or human reviewer comment, CodeCrab can perform a read-only "deep dive." It investigates the surrounding code, connects it to project context, and can generate a verified fix. The fix is tested against the local test suite (e.g., pytest, npm test) before the developer chooses to apply it.
3. Pre-Push Local Change Analysis: Before creating a Pull Request, developers can review their local, uncommitted, or unstaged changes. This feature catches bugs, anti-patterns, and regressions early in the development cycle, ensuring higher-quality initial PR submissions.
4. Adaptive Codebase Profile & Skill Integration: The application learns repository-specific patterns and rules over time, building a custom review profile. It allows engineers to integrate and orchestrate their own local review scripts ("skills") alongside built-in CodeCrab agents and commercial AI CLI tools, enabling limitless customization.

Problems Solved

1. Pain Point: The security and compliance risk of uploading proprietary or sensitive source code to third-party SaaS code review bots. CodeCrab solves this with its zero-code-upload architecture.
2. Pain Point: The inefficiency and context loss of switching between an AI chatbot, a diff viewer, and a local terminal to understand and fix code issues. CodeCrab unifies investigation, review, and fixing in a single desktop interface.
3. Target Audience: Software engineers and engineering teams in regulated industries (finance, healthcare), enterprises with strict data governance, open-source maintainers, and privacy-conscious developers who already use AI coding assistants like Claude Code or Cursor.
4. Use Cases: Rapidly reviewing teammate PRs in a private repository; conducting a final quality check on AI-generated code before pushing; deeply investigating a vague code review comment to understand the root cause and correct fix.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros:
     • Unmatched Privacy & Security: Code never leaves the local machine, making it compliant with strict corporate IT policies.
     • Full Control & Transparency: Developers use their own AI API keys/subscriptions and see real-time CLI output in the Live Execution Console.
     • Deep Workflow Integration: Leverages existing local tooling (Git, AI CLIs, test suites) without requiring disruptive cloud OAuth permissions.
     • Economic Efficiency: Avoids per-review SaaS fees by reusing existing AI tool subscriptions.
   • Cons:
     • Local Resource Dependency: Review quality and speed are dependent on the power of the local machine and the capabilities of the connected AI CLI tools.
     • Setup Complexity: Requires initial configuration of local AI CLIs and Git credentials, which may present a barrier compared to cloud-based "sign-in-with-GitHub" solutions.
     • Platform Limitation: Currently only available as a desktop application for Windows (as of Beta v0.1.10), excluding macOS and Linux users.

2. Key Alternatives & Differentiation:

   • Cloud SaaS Bots (e.g., SonarCloud, Codacy, PullRequest.ai): These require full code upload to their servers for analysis. CodeCrab differentiates on privacy, data sovereignty, and the ability to run local tests. It also avoids requiring broad GitHub OAuth permissions at the organization level.
   • IDE-Based AI Assistants (e.g., GitHub Copilot, Cursor, Claude Code): These are excellent at code generation and inline suggestions but lack a dedicated, structured workflow for holistic Pull Request review. CodeCrab complements them by providing a diff-centric review pane and orchestration specifically for the review phase.
   • Open-Source Static Analysis Tools (e.g., Semgrep, CodeQL): These run locally but are rule-based. CodeCrab differentiates by integrating adaptive, LLM-powered analysis that can understand context and intent, while still allowing the integration of such tools as custom local "skills."

Frequently Asked Questions (FAQ)

1. Does CodeCrab upload my code to the cloud? No. CodeCrab is a 100% local-first desktop application. Your source code, diffs, and AI analysis never leave your laptop. It only communicates externally to use the AI APIs (e.g., Anthropic's, OpenAI's) that you configure with your own keys.
2. What AI models does CodeCrab support? CodeCrab does not provide its own AI models. Instead, it orchestrates local AI command-line tools you already have, such as Claude Code, Cursor's AI, or the GitHub Copilot CLI. You bring your own subscriptions and API keys.
3. Can CodeCrab automatically approve or comment on my GitHub Pull Requests? No. CodeCrab is designed as a read-only assistant for the reviewer or author. It does not have permission to post comments or approve PRs on its own. Its "assisted fix" feature applies changes only to your local working branch after your verification.
4. How does CodeCrab compare to using ChatGPT for code review? While you could manually paste a diff into ChatGPT, CodeCrab automates the entire workflow: fetching the PR, providing full repository context to the AI, parsing the response, mapping findings to the exact lines of code, and enabling deep-dive investigations—all within a secure, local environment without copy-pasting.
5. Is CodeCrab free? The application is currently in a Free Public Beta. The long-term pricing model is not yet defined, but the core premise is to avoid per-review fees by relying on the user's own AI infrastructure costs.