Diffblue today announced the general availability of its Diffblue Testing Agent, an autonomous regression test generator designed to work alongside AI coding platforms such as GitHub Copilot and Claude Code. The agent automatically produces verified unit tests across entire codebases without developer intervention, addressing a longstanding challenge in enterprise software engineering: generating comprehensive, trustworthy test coverage at scale.

Closing the Gap in AI-Assisted Development

AI coding agents have transformed how developers write software, but creating robust regression tests has remained labor-intensive. Developers often spend hours iterating with AI coding assistants, manually verifying output, and coaxing adequate coverage—frequently achieving suboptimal results.

Diffblue’s Testing Agent introduces an orchestration and verification layer that autonomously scopes codebases, generates tests at the method and class level, validates compilation, verifies test results, rolls back failed tests, and even prepares pull requests—all without manual intervention.

“Our benchmark data shows that the developer effort for driving even the best AI coding agents reaches unaffordable levels quickly,” said Dr. Peter Schrammel, co-founder and CTO of Diffblue. “The Diffblue Testing Agent achieves 80%+ coverage autonomously—the difference between an AI experiment and an AI-enabled engineering workforce.”

Benchmark Performance

In tests across eight real-world Java projects, Diffblue delivered:

• 81% average line coverage, compared to 32% achieved by an AI-savvy developer using only Claude Code
• 61% mutation coverage, surpassing typical enterprise standards
• 2.5x higher coverage than iterative human-assisted AI workflows

The benchmarks highlight the platform’s ability to scale efficiently across hundreds or thousands of classes in a single run, turning AI coding tools from experimental assistants into reliable engineering collaborators.

How It Works

The Diffblue Testing Agent integrates with existing AI coding platforms, delegating test generation to the underlying agent while orchestrating:

• Coverage analysis and build system configuration
• Test plan creation and parallelized generation
• Verification and rollback of failing tests
• Project clean-up and pull request preparation

Initially available for Java and Python, Diffblue plans to expand to additional platforms and software quality domains, including test quality assessment, code review automation, large-scale refactoring, and requirements-driven test generation.

Enterprise Impact

By automating regression test creation, Diffblue enables engineering teams to:

• Accelerate modern application development
• Improve software quality and reliability
• Reduce developer toil and resource costs
• Confidently scale AI-assisted workflows across the organization

The Diffblue Agents platform, founded by researchers from the University of Oxford and backed by IP Group, Albion, Parkwalk, and Citi, is now available for enterprise evaluation.

Get in touch with our MarTech Experts.