GitHub says 67.3 percent of problems flagged by its Code Quality service in its own engineering teams get resolved before pull requests are merged, according to an announcement from the company. The service, now generally available on GitHub Enterprise Cloud and GitHub Team, pairs CodeQL analysis with AI-powered identification of maintainability and reliability issues, then offers Copilot Autofix to recommend changes during code review. GitHub presents its internal figure as an early indicator rather than a performance guarantee.
The product spent time in public preview starting in October 2025, during which more than 10,000 enterprises tested the service, GitHub reports. The general availability version adds company-wide activation, dashboards displaying maintainability and reliability metrics, test coverage data, and rulesets for enforcing quality gates. The system works at two points in the development cycle: on pull requests, where CodeQL surfaces quality problems in context and tracks whether changes impact test coverage, and on the default branch, where it maps existing quality debt throughout a repository. Teams can apply rulesets to mandate coverage or other benchmarks, with an evaluate mode designed to support phased deployment.
GitHub frames the release as a response to AI-driven code generation, noting that while AI-assisted development tools boost the volume of code created, the responsibility for evaluating design and long-term maintainability stays with engineering teams. The announcement states that "Code Quality helps teams ship code they trust," positioning the service as a control mechanism for maintainability, coverage and reliability that complements CodeQL's existing security checks. The company acknowledges that Code Quality can spot and rank problems, but the choice to accept a proposed fix remains with developers. GitHub's approach builds on its established use of CodeQL and Copilot Autofix for security vulnerabilities, a model Microsoft is also deploying for Azure Repos, where static analysis combines with AI-generated remediation—though that implementation notes generated fixes still require standard testing and human oversight because they can be partial or create unexpected side effects.
Code Quality is sold as a standalone paid product, separate from GitHub Advanced Security, with a base cost of $10 per active committer monthly. An active committer is tallied once per organization when they've pushed to an activated repository in the prior 90 days. AI-powered detection and Copilot Autofix carry usage-based fees, while deterministic CodeQL scans consume GitHub Actions compute on both GitHub-hosted and self-hosted runners. For preview customers, existing setups continue under their current GitHub agreement, but organizations must audit where Code Quality is enabled to avoid unexpected charges. The service launched on GitHub Enterprise Cloud and GitHub Team, though it's absent from GitHub Enterprise Server at release. GitLab is pursuing a parallel path with Duo Code Review, which examines merge requests using repository, pipeline, security and compliance context and can produce structured inline feedback at a flat $0.25 per agentic review regardless of change size, available across GitLab.com, Dedicated and supported self-managed installations. Atlassian's Rovo Dev connects code review to work-management context, assessing pull requests in Bitbucket and GitHub for quality, security and performance while cross-referencing Jira acceptance criteria and project background to verify implementation aligns with intended work; Atlassian reports its internal Rovo Dev Code Reviewer cut pull request cycle time by 30.8 percent, though that outcome derives from the company's own engineering setup and isn't an independent benchmark. A Reddit discussion responding to the launch centered on access and pricing, with users noting the fee is tied to committers rather than reviewers and questioning whether organization-wide visibility and automated fixes warrant another per-user platform charge. The broader trend suggests that as generative AI accelerates code output, platforms are racing to layer automated review and remediation on top of human judgment, betting that the marginal cost of machine checks will prove smaller than the technical debt left unaddressed.