Kodus

Visit Website

Introduction

Kodus is an open-source AI code review platform built around an agent called Kody. It reviews pull requests directly in GitHub, GitLab, Bitbucket, Azure DevOps and Forgejo, and you can run it on your own servers or use Kodus Cloud.

Its central idea is model freedom. Kodus is model-agnostic: you connect your own provider credentials and choose the models behind reviews, whether OpenAI, Anthropic, Google Gemini, Vertex AI, Novita or any OpenAI-compatible endpoint, including open-weight models such as Llama, GLM and Kimi. Billing stays in your own provider account, and Kodus advertises zero markup on LLM costs. A token usage view shows consumption across reviews so model spend stays predictable as adoption grows.

Kody does more than summarise diffs. It reviews code with context, flags risks by severity and suggests concrete fixes in the pull request; the README shows it catching a critical insecure direct object reference (IDOR) where an organisation identifier was trusted from user input. Kody Rules let teams write review instructions in plain language and apply them across organisations, repositories, paths or specific review scopes, enforcing architecture decisions, security expectations, testing practices and local conventions without reviewers repeating themselves.

Kodus also follows up. It tracks unimplemented suggestions from closed pull requests, lets you manage them by status, severity, category and repository, and resolves them when the fix appears in a later PR. Cockpit measures review effectiveness, rule health, repository health and delivery metrics across the engineering workflow.

For local work there is a CLI. kodus review reviews your working changes, --staged restricts it to staged files, and --prompt-only produces output meant for a coding agent to act on. The same reviews can run in CI pipelines.

On privacy, the project states that source code is not used to train models, data is encrypted in transit and at rest, and self-hosted runners are supported. Self-hosted instances send one anonymous daily heartbeat with aggregated counters and no code or identifiers, which you can disable with KODUS_TELEMETRY_DISABLED=true. The self-hosting guide suggests at least 2 CPU cores, 8 GB of RAM and 60 GB of free disk.

A sensible rollout for any AI reviewer is to enable it on two or three active repositories, ask the developers who receive its comments to rate them for a fortnight, and only then widen it. Track how many comments lead to a code change: that ratio says more about value than any benchmark chart.

The source code is at kodustech/kodus-ai on GitHub under the AGPL-3.0 (outside its enterprise code) licence, with about 1,600 stars; the most recent commit was on 11 November 2026.

Self-hosting makes sense when code cannot leave your network or when you want to run open-weight models on your own GPUs; Kodus Cloud is simpler if neither applies.

Features

1. Multi-platform PR reviews: GitHub, GitLab, Bitbucket, Azure DevOps and Forgejo.
2. Bring your own model: Any OpenAI-compatible endpoint.
3. Zero LLM markup: Pay providers directly.
4. Kody Rules: Plain-language review rules with scopes.
5. Severity and fixes: Concrete suggestions in the PR.
6. Suggestion tracking: Follows up on unresolved suggestions.
7. Cockpit metrics: Review and delivery insights.
8. CLI: Local and agent-friendly reviews.
9. Self-hosting: Run on your own infrastructure.
10. Telemetry opt-out: Single environment variable.

Pros

1. Cost control: No hidden model margins.
2. Platform breadth: Supports less common hosts like Forgejo.
3. Free to use: Open source under the AGPL-3.0 (outside its enterprise code) licence, so the core tool costs nothing to use.
4. Actively maintained: The repository had commits as recent as 11 November 2026.

Cons

1. Self-hosting effort: Needs a modest server and maintenance.
2. Model quality varies: Results depend on the model you pick.
3. AGPL core: Licence obligations for modified network use.

Usage and Links (Example)

kodus review            # review working-tree changes
kodus review --staged   # only staged changes
kodus review --prompt-only   # output for a coding agent
kodus --help

Self-host from the kodus-ai repository, or create a Kodus Cloud account and connect your Git provider.

Resources:

• Kodus website
• GitHub repository

Pricing

The Community plan self-hosts for free. Kodus Cloud offers a 14-day free trial with up to 35 PR reviews and no credit card; paid plans are on the pricing page, and LLM costs are paid directly to your provider.

Recommended for?

1. Teams that must keep code and model choice in-house.
2. Organisations on GitLab, Bitbucket, Azure DevOps or Forgejo.
3. Cost-conscious teams using open-weight models.

Conclusion

Kodus is one of the most flexible open-source AI reviewers, with real self-hosting and no model lock-in.

Pair Kody Rules with your existing linters and keep humans as the final approvers.

On CheckOps, Kodus is filed under the Release & CI gates and Unit testing stages. Related tools listed here include CodeRabbit, PR-Agent, Greptile, Sourcery, CodeAnt AI and Cursor Bugbot. Which one fits best depends mostly on your stack, your team’s size and skills, and whether you prefer to run tools yourself or pay for a managed service. This page reflects the official website, documentation and repository as they stood on 11 November 2026.