Overview:
Kodus is an open-source AI code review tool that gives teams full control over model selection and LLM costs. It connects directly to existing Git workflows via pull requests on GitHub, GitLab, Bitbucket, and Azure Repos. The tool is designed for development teams that want AI-assisted code review without opaque pricing or vendor lock-in on AI models. It supports a bring-your-own-key model, allowing teams to pay providers like Claude or GPT directly with zero markup.
Core Features:
Model Agnostic: Supports Claude, GPT-5, Gemini, Llama, GLM, Kimi, or any OpenAI-compatible endpoint for code review.
Zero Markup on LLM Costs: Teams pay model providers directly, with no hidden multipliers or extra charges.
Custom Review Rules (Kody Rules): Define review criteria in plain language; the Community edition supports up to 10 rules.
Kody Learnings and Memory: The system adapts to the team's architecture, standards, and workflow over time.
Native Git Workflow: Operates directly within pull requests across GitHub, GitLab, Bitbucket, and Azure Repos.
CLI + CI/CD Ready: Can run reviews locally or within CI/CD pipelines.
Use Cases:
Teams wanting to use specific AI models: Developers can choose Claude, GPT, Llama, or other models for code review without being tied to a single provider.
Cost-conscious development teams: Organizations that want to control and audit their AI spending by paying model providers directly.
Self-hosters requiring data control: Teams that need to keep source code on their own infrastructure, supported by self-hosted runners and on-premise deployment options.
PR-based code review workflows: Teams already using GitHub, GitLab, Bitbucket, or Azure Repos can integrate AI reviews directly into their existing pull request process.
Why It Matters:
Kodus addresses a common friction point in AI-powered development tools: cost opacity and model lock-in. By allowing teams to bring their own API keys and pay providers directly, the project removes hidden multipliers. Its support for multiple AI models and self-hosted deployment gives technical teams flexibility in tooling and data control. The Community edition provides a free, self-hostable option with unlimited PR usage and core features like custom rules and Kody memory.