Proval is a privacy-centric, self-hosted tool that integrates with GitLab, Forgejo, and GitHub to automate code reviews using AI models of your choice.
Imagine this: before you show the code you’ve carefully crafted to your colleagues, an AI reviews it thoroughly first. A friendly AI that offers advice like, “There’s a typo here,” or “This code could be optimized.” But what if you are hesitant to send your company’s core source code to an external entity? Recently, an interesting tool has emerged to solve exactly this dilemma: Proval.
Why does this matter?
In software development, “Code Review”—the process of examining a peer’s code to find errors and improve quality—is essential. However, having a human review every line of code is time-consuming and energy-draining. While services where AI performs this task are increasing, the security concern that a company’s critical code might be transmitted to an external AI server remains.
Proval targets this point precisely. By using a “self-hosted” approach—where software is installed and operated directly on your own server rather than via an external service—it ensures that code data never leaves your environment, providing peace of mind to enterprises and individual developers alike where security is critical. Source 1
In simple terms, while existing AI code review tools are like ordering food from a communal “cloud” kitchen, Proval is like hiring a dedicated chef directly for your company’s kitchen. Since data never leaves your company’s server, you can reduce the worry of confidential information leaks.
How does it work?
The core of Proval is that you are free to choose the “chef” that suits your taste.
- Model selection at your discretion: Proval’s greatest strength is its “Bring your own model” strategy. Users can directly connect their preferred AI models to their server using tools like Ollama or llama.cpp. Source 1 Source 8
- Easy installation: To lower the technical barrier to entry, it can be installed using a single Docker image, which packages the environment required to run the software. Source 6
- Seamless integration: It currently supports smooth integration with popular development platforms such as GitLab, Forgejo, and GitHub. Source 2 Source 8
What is the current status?
Proval is currently in the early stages of its journey. It was created by a developer who wanted to automate code reviews in a self-hosted environment, and some features are still raw or in need of refinement. Source 2 Source 3
It is a tool optimized for users who manage their own servers in home-lab environments, teams that must work in air-gapped networks with restricted external internet access, and infrastructure teams that prioritize security above all else. Source 4
The future ahead
Moving forward, Proval is expected to improve, allowing users to integrate a wider variety of AI models more freely and making it easier to install and operate in complex environments. In that it enables the enhancement of development productivity using the latest AI technology even in air-gapped environments, it will likely become a powerful option for security-conscious companies.
However, since it is an early version, it is recommended to keep a close eye on the project’s updates before considering adoption. If you are a developer who manages your own servers, why not install it in a test environment right now and build your own reliable “AI security guard”?
References
- Proval - Self-hosted AI code review infrastructure: https://proval.app/
- GitHub - seoes/proval: Self-Hosted LLM Code Review Agent with…: https://github.com/seoes/proval
- Show HN: Proval – Self-hosted code review agent for GitLab, Forgejo, and…: https://hn.today/s/show-hn-proval-self-hosted-code-review-agent-for-gitlab-forgejo-and
- Show HN: Proval – Self-hosted code review agent for GitLab, Forgejo, and…: https://modernorange.io/item/49465821
-
seoes/proval — GitHub trending stats & insights Trendshift: https://trendshift.io/repositories/95306 -
Show HN: Proval – Self-hosted code review agent for GitLab, Forgejo, and GitHub Hacker News: https://news.ycombinator.com/item?id=49465821
- It performs all reviews strictly on an external cloud
- Users can manually select and install their own AI models
- A paid subscription is mandatory
- GitLab, Forgejo, and GitHub
- Only GitHub
- GitLab and Slack
- Environments where an internet connection is mandatory
- Teams operating within air-gapped or on-premises infrastructure
- Teams that only want to use cloud services