Lightrun has launched a pull request review tool called Runtime Aware PR Verifier, designed to test proposed code changes against live production behaviour before deployment.
The tool simulates how a proposed code change would behave on the production code paths it affects, using live runtime data rather than relying only on tests, static analysis or earlier production patterns. Each pull request receives a risk score ranging from risky to safe based on its performance across live execution paths, dependency interactions and real traffic.
The launch reflects a growing focus in software development on the limits of conventional code review as teams produce larger volumes of software changes, often with the help of AI coding tools. In that environment, companies are looking for ways to spot regressions and implementation errors before code reaches production systems.
The verifier also reviews the original ticket associated with a pull request and checks how the proposed code would affect live runtime activity. This is intended to test whether requested functionality and edge cases are reflected in the implementation, while reducing false positives that can come from isolated test environments.
Workflow focus
According to Lightrun, the product delivers its findings within GitHub, GitLab and Bitbucket. That keeps review output inside the existing development workflow rather than requiring engineers to switch between separate tools.
Lightrun says the tool could lead to fewer production incidents after merges, lower review and quality assurance costs, improved deployment frequency, reduced mean time to repair, and lower cloud and data costs. Those claims reflect the commercial pressure on software teams to catch defects earlier, when they are usually cheaper to fix.
The company positions the product as addressing a gap between traditional testing and production reality. Test suites and static code analysis can identify many faults before release, but they often struggle to reproduce the exact conditions, dependencies and traffic patterns of live systems.
That problem has become more visible as AI-assisted coding increases the number of pull requests engineering teams must examine. More code submitted for review can raise the risk that subtle flaws slip through standard processes even when formal tests pass.
Or Golan, R&D Lead of Lightrun's AI Labs, linked the launch to that shift in development practice.
"AI coding agents have removed the bottleneck on writing code; a backend engineer using Copilot or Claude Code can produce 10 times more PRs than they could two years ago," said Or Golan, R&D Lead, Lightrun.
He said the main concern is not only the volume of code changes, but the type of defect that can survive existing checks and only become visible after release.
"However, with AI comes bugs, and the most dangerous bugs are those that pass code review, clear every test, and then quietly break existing functionality after deployment. Our Runtime Aware PR Verifier identifies these hidden issues by testing how the pull requests would actually behave in production, even when traditional testing shows no problems," Golan said.
Company background
Lightrun describes itself as an AI-focused reliability engineering platform for development and site reliability teams. It says its software is used to prevent and address software issues across the software development lifecycle by generating telemetry at line level and validating fixes in runtime environments.
The company says its customers include AT&T, Citi, Microsoft, Salesforce, UnitedHealth Group, SAP, ICE/NYSE, ADP, HPE and Booking Holdings. It also says it has raised USD $110 million from investors including Accel and Insight Partners.
The release of a tool centred on runtime verification shows how software vendors are trying to bring code review closer to production conditions without waiting for deployment. For engineering teams dealing with higher volumes of AI-assisted changes, the approach aims to find failures in the narrow space between passing tests and real-world system behaviour.
Lightrun says the verifier catches failures missed by tests, static analysis and AI reviewers.