
We launched our code review agent, Qodo Merge (formerly PR Agent) in 2023, as one of the first AI-driven code review tools, helping to automate critical parts of the code review process, from generating PR descriptions, review guide, code suggestions and more.
More than a year later, Qodo Merge has matured into a robust code review agent, with 14 workflows (called via commands like /describe, /review, and /improve) and a suite of features designed to streamline code reviews. Yet, with every leap forward, new challenges emerge, from redundant feedback to a flood of low-priority suggestions.
In the latest batch of releases to Qodo Merge, we’re addressing new challenges that have become inherent in AI-assisted coding—optimizing feedback relevance, enhancing adaptability to team-specific coding practices, and ensuring a seamless developer experience.
With these capabilities, we’re excited to announce Qodo Merge 1.0. New features and capabilities include:
- Focus on problems mode: prioritizing critical issues like bugs, security flaws, and maintainability while filtering out stylistic noise to keep the review process efficient and impactful.
- Dynamic learning: by analyzing patterns in accepted acceptions, Qodo Merge maintains and uses a best practices wiki to refine future suggestions, ensuring that feedback aligns with the team’s evolving coding standards.
- Real-time ticket context: Seamlessly integrates ticket requirements and details into the code review, ensuring compliance with requirements during reviews.
- Actionable feedback with /implement command: convert code review discussions and feedback into actionable, ready-to-commit code changes.
Qodo Merge 1.0 is tailored to meet the evolving needs of developers, making the code review process more precise, adaptive, and efficient. Let’s take a deeper look at the new capabilities in Qodo Merge 1.0:
From noise to signal reduction
Anyone who’s used AI code review tools has likely encountered a similar scenario: You open a PR for a minor bug fix in your authentication service, and suddenly you’re bombarded with suggestions about naming conventions, optional chaining patterns, and “did you consider using a different design pattern?” Meanwhile, the critical edge cases in your error handling logic go unnoticed.
Take this example:
async function authenticateUser(credentials) { try { const user = await getUserByEmail(credentials.email); if (!user) return null; const isValid = await validatePassword(credentials.password, user.hash); return isValid ? user : null; } catch (err) { console.error(err); return null; } }
You might be used to getting code suggestions like:
- “Consider renaming ‘credentials’ to ‘userCredentials’ for clarity”
- “You could use optional chaining here: credentials?.email”
- “Have you considered using the Strategy pattern for different auth methods?”
While these suggestions aren’t wrong per se, they miss more critical issues like:
- Silent failure in error handling
- No distinction between auth failures and system errors
- Potential timing attacks in the password comparison
Users of early versions of Qodo Merge encountered this, hindering their ability to promptly address critical issues. This led us to develop a new default focus mode that prioritizes the issues that matter most. With a novel approach to reducing noise, this mode brings attention to high-impact problems like security vulnerabilities and error resilience, enabling developers to address critical concerns efficiently and elevate the overall quality of their code.
Having rolled out this feature to some users during the pre-release phase, we saw an increase in acceptance rates by nearly 50%, demonstrating its effectiveness in guiding developers toward meaningful improvements.
AI that adapts to your best practices
AI tools in code reviews often provide advice that feels disconnected from your team’s specific needs. For example, a recommendation to “use a third-party library for this functionality” could ignore your team’s policy of minimizing external dependencies for security and maintenance reasons. These generic suggestions often miss the nuanced decisions and trade-offs that are integral to your codebase and workflows.
Qodo Merge addresses this gap by offering customized, context-aware feedback tailored specifically to your team’s workflows and coding practices. It learns from your team’s accepted code suggestions, using LLM-powered flows to automatically detect and analyze these patterns. By tracking accepted suggestions over time, it builds a dynamic best practices wiki (detailed here) and continuously refines its recommendations.
Through this process, Qodo Merge creates a living repository of coding standards (learn more) that evolves with your team, ensuring feedback remains relevant, actionable, and aligned with your priorities.
Context-aware reviews with ticket compliance
Qodo Merge’s ticket context fetching brings relevant information directly into the pull request. As soon as you open a PR, Qodo Merge automatically identifies the associated ticket—whether it’s in GitHub Issues, Jira Cloud, or Jira Server—and surfaces key details like acceptance criteria, ticket descriptions, comments, and dependencies.
Let’s say you’re reviewing a PR that implements a new onboarding flow. Qodo Merge pulls in the ticket’s requirements, showing that the design team specified a step-by-step progress bar as a critical component. Without this context, a reviewer might overlook its absence, approving changes that don’t fully meet the project’s goals. With Qodo Merge, these gaps are eliminated—everything you need is right where you need it.
This rich ticket context doesn’t just aid human reviewers—it also powers Qodo Merge’s automated review capabilities. When you run /describe or /review commands, Qodo Merge leverages the ticket information to provide more insightful analysis of your code changes. By understanding the original requirements and intent behind the modifications, the AI can offer more relevant and contextual feedback.
Additionally, Qodo Merge performs automated ticket compliance analysis, evaluating how well the PR fulfills its intended purpose. Each PR receives a compliance rating—Fully compliant, Partially compliant, or Not compliant—based on how closely the changes align with the requirements specified in the associated ticket. This ensures that no critical requirements slip through the cracks during the review process.
Transform review feedback into actionable code
Qodo Merge’s new /implement command (supported for Github, GitLab and Bitbucket Cloud and Bitbucket Server) directly implements code changes based on reviewer feedback.
Qodo Merge uses its understanding of your codebase and best practices to automatically generate precise code changes from review comments. For example, if a reviewer suggests, “Wrap code with try-except” the /implement command can directly apply the change by identifying the relevant code block, inserting a try-except construct, and ensuring proper exception handling is in place.
With the features we’ve introduced in version 1.0, we’re aiming to build systems that understand the nuanced context of your codebase, team practices, and business requirements. The result is a more intelligent review process that adapts to your team’s needs while maintaining high standards for code quality.
What’s next for Qodo Merge
Looking ahead, we’ll continue to bring developer-focused enhancements designed to tackle complex challenges in code reviews. Advancements in automated learning will help tailor recommendations based on your team’s accepted practices, while updates to the /review command will deliver more precise, actionable feedback. With expanded analytics, Qodo Merge will deliver the flexibility and insights developers need to maintain high-quality, secure code at scale.