.svg)
Since launching in December, Codacy’s AI Reviewer has been helping engineering teams get more signal and less noise out of their Pull Request reviews. Today, we’re sharing three major updates that enhance its functionality, along with an important change for teams currently using AI-enhanced comments.
What's New
Custom instructions
Every codebase is different, and now you can tell the AI Reviewer exactly what to look for in yours. Add a review.md file to your repository to guide its behavior and tune the quality of its output, from coding conventions to the type of feedback that matters most to your team.
Not sure where to start? We provide a ready-to-use prompt for your coding agent to generate that file from scratch.
Analyze this repository and generate a concise AI PR reviewer instruction file in Markdown. The file should give a PR reviewer the essential context that won't be visible in a diff. Cover:
- **Purpose**: What this repo does in 1-2 sentences.
- **Architecture**: High-level structure, patterns used (e.g. MVC, event-driven, monorepo, etc.).
- **Folder structure**: Key directories and what lives in each. Skip obvious ones.
- **Stack**: Languages, frameworks, major libraries, and their roles.
- **Testing**: Framework used, where tests live, what's expected to be tested.
- **Code style & conventions**: Naming, file organization, formatting tools, any patterns enforced.
- **PR-specific rules**: Branch strategy, what a PR should/shouldn't include, migration rules, etc.
- **Common pitfalls**: Things reviewers should flag — anti-patterns, known gotchas, areas that break easily.
- **Out of scope**: Anything reviewers should explicitly ignore or not enforce.
Rules for the output:
- Output only the Markdown content, no preamble or explanation.
- Be concise. Every line should earn its place — no filler.
- Use short sections with bullet points. Avoid long prose.
- If something is not applicable or not inferable, omit the section entirely.
Test reviews
The AI Reviewer now analyzes your changes and leaves suggestions for missing tests. When it spots gaps in test coverage, it also provides a prompt proposal you can hand straight to your coding agent to implement. This turns a review comment into an actionable next step.
On-demand reviews
A new Run Reviewer button on the Pull Request gives you control over when the AI Reviewer runs. You can choose between two modes:
- Manual mode: Manually trigger a review whenever you’re ready.
- Automatic mode (first review): The AI Reviewer runs once when the PR is opened. From there, you can re-trigger it manually as you commit new changes, without generating additional noise.
AI-enhanced comments are being replaced by the AI Reviewer
If your GitHub organization has been using AI-enhanced comments, here’s what you need to know: on March 17, 2026, AI-enhanced comments will be deprecated and automatically replaced by the AI Reviewer.
The AI Reviewer takes a step forward and adds deeper insights. Where AI-enhanced comments focused on individual issues in isolation, the AI Reviewer analyzes the intent of your entire Pull Request, drawing context from the files changed, PR metadata, and optionally Jira tickets associated with the PR. This translates to more accurate feedback, better fix suggestions, and significantly less noise.
As with AI-enhanced comments, none of your data is used to train any AI models.
What do you need to do?
Nothing at all. On March 17, all GitHub repositories with AI-enhanced comments enabled will be automatically switched to the AI Reviewer. Repositories where AI-enhanced comments are not enabled won’t be affected.
If you’d rather not wait, you can enable the AI Reviewer in your organization and repository integration settings now. Just make sure to manually disable AI-enhanced comments at the same time to avoid receiving duplicate comments on your Pull Requests.
Note: the AI Reviewer is currently only available on GitHub, for all Team and Business plans.
Share your feedback
We’re actively improving the AI Reviewer and would love to hear how it’s working for you. Fill out this short form to share your feedback.
