Metabob

Metabob is an AI-powered code review tool that detects, explains, and fixes coding problems in software created by both humans and AI. It utilizes graph neural networks (GNNs) and large language models (LLMs) to provide comprehensive, context-sensitive analysis of code, helping developers improve code quality, reliability, and security.

Key features:
- Detection of Logical Problems: Identifies complex issues such as race conditions and unhandled edge cases that traditional static analysis tools might miss.
- Contextual Understanding: Uses GNNs to understand the semantic and relational markers in the code, providing a more complete representation.
- Automated Fixes: Generates context-sensitive problem explanations and resolutions using LLMs.
- Integration with Pull Requests: Analyzes each pull request to improve code quality and security without requiring CI setup.
- Enhanced Security: Scans code for security vulnerabilities and provides fixes, ensuring that software is secure before deployment.
- Increased Productivity: Saves significant debugging time by automatically identifying and fixing issues, thus increasing developer productivity.
- Comprehensive Analysis: Analyzes the entire codebase to detect and recommend fixes for critical errors.
- On-premises Deployment: Offers a self-hosted version for organizations that prefer to keep their code and analysis in-house.

How it works:
1. GNNs detect and classify problematic code with contextual understanding.
2. Problematic code along with enriched context is stored in Metabob's backend.
3. The stored information is passed to an integrated LLM.
4. The LLM generates a context-sensitive problem explanation and resolution.

User Interaction:
- Users can interact with Metabob through a web app or directly within their integrated development environments (IDEs) such as VS Code.
- Once integrated, Metabob automatically analyzes code and provides recommendations without requiring user prompts.

Integrations:
GitHub, Bitbucket, GitLab, VS Code

Use of AI:
Metabob combines GNNs and LLMs in its analysis process. The GNNs provide a deep understanding of the code's structure and semantics, while the LLMs generate explanations and fixes based on billions of parameters and an attention mechanism.

Target users:
- Developers looking to improve their code quality and productivity
- Development teams and large enterprises aiming to enhance their software security and reliability

How to access:
Metabob is available as a web app, IDE plugin (VS Code), and self-hosted solution.