At Quarkus, we welcome tools that help developers become more productive — including AI tools such as Large Language Models (LLMs) and agents like Claude Code, Codex, ChatGPT, GitHub Copilot, and others.
However, recent patterns of use have led to increased moderation burden, low-value contributions, and reduced community signal. To ensure a healthy and productive community, the following expectations apply to all contributions (issues, pull requests, comments, discussions, and other project interactions).
- You may use agents/LLMs to help you write better, but not to post more.
- AI may be used to assist your development — for example drafting code, writing documentation, or proposing fixes — as long as you understand, validate, and take responsibility for the results.
- You should only submit contributions (PRs, comments, discussions, issues) that reflect your own understanding and intent, not what an agent/LLM "spit out."
- This means that when the contribution is scrutinized by a maintainer, you need to be ready to argue about all aspects of it; simply responding that the contribution is what the LLM provided is completely unacceptable
- Submitting code, tests, comments, discussions or issues that appear to be copied directly from an AI with little or no human oversight is not acceptable.
- Posting large volumes of low-effort suggestions, vague issues, or links with no context — even if technically accurate — is considered spam.
- Submitting AI-generated tests that do not validate actual behavior or meaningfully cover functionality is not helpful and will be rejected.
- Using bots, agents, or automated tools to open PRs, file issues, or post content without human authorship and responsibility is not allowed.
As a general rule, the actual PR and its metadata should be similar to what the developer would have come up with had they not been using AI tools. For example, when writing PR descriptions, developers tend to make it clear what the PR does and don't pack the description with useless details or superfluous formatting and emojis. Furthermore, developers usually try to introduce focused tests that follow the project's existing test philosophy instead of dumping a bunch of new tests for everything that could conceivably be tested concerning the updated code.
- We may close issues, PRs, or discussions that violate this policy without detailed explanation.
- Repeated violations may result in temporary or permanent restrictions from participating in the project.
If you're unsure whether your use of agents/LLMs is acceptable — ask! We're happy to help contributors learn how to use AI tools effectively without creating noise.
This isn’t about banning AI — it’s about keeping Quarkus collaborative, human-driven, and focused on quality.