kodustech/kodus-ai

What it solves

Kodus automates the code review process for pull requests, reducing the manual effort required by human reviewers to catch bugs, security risks, and architectural inconsistencies. It helps teams maintain high code quality and enforce specific standards across multiple Git providers without relying on repetitive manual feedback.

How it works

Kodus integrates directly into the Git workflow (GitHub, GitLab, Bitbucket, Azure DevOps, and Forgejo) via webhooks. It uses Large Language Models (LLMs) to analyze code diffs and provide concrete fixes and risk assessments directly in the PR. Users can define "Kody Rules" in plain language to guide the AI on specific architecture or security requirements. The system is model-agnostic, allowing users to bring their own API keys (BYOK) for providers like OpenAI, Anthropic, or Google Gemini, or use self-hosted models.

Who it’s for

Engineering teams and developers who want to automate code reviews, enforce coding standards across a large organization, or maintain strict privacy by self-hosting their review infrastructure.

Highlights

• Model Agnostic: Supports various LLMs including Claude, GPT-5, Gemini, Llama, and any OpenAI-compatible endpoint.
• Customizable Rules: Allows teams to define plain-language review instructions applied to specific repositories or paths.
• Self-Hostable: Can be deployed on private servers via Docker to ensure source code privacy.
• Native Git Integration: Works directly within PRs across five major Git platforms.
• CLI & CI/CD Support: Enables running AI reviews locally from the terminal or within automated pipelines.
• Issue Tracking: Automatically tracks unimplemented suggestions from closed PRs and resolves them when fixes appear in future PRs.