Diffblue

What Diffblue does

Diffblue Cover is an AI agent for automating the generation, maintenance, and management of Java unit tests, enabling developers to focus on building high-quality code. The Diffblue Testing Agent orchestrates AI coding tools to create comprehensive, high-quality test coverage with minimal developer intervention. It is designed to modernize legacy code with confidence and is significantly faster than manual test writing or other coding assistants. The platform uses reinforcement learning to generate code that is guaranteed to run, compile, and be correct, operating on-premise to ensure code security and privacy. Diffblue offers solutions for individual developers and enterprise-scale deployments, with outcome-based pricing tied to net new lines of coverage added.

Products

• Diffblue CoverAn AI agent for automating the generation, maintenance, and management of Java unit tests. It automates unit testing, empowering dev teams to deliver high-quality software faster, efficiently, and more reliably than coding assistants or writing tests manually. It writes comprehensive, human-readable unit tests, updates and maintains them after each code change, documents code, provides regression detection, enables rapid coverage increases, and offers feedback on best testability and coding practices.
• Diffblue Testing AgentAn autonomous regression unit test generation tool for Java and Python projects. It orchestrates workflows through existing AI coding agent platforms (like GitHub Copilot and Claude Code), handling scoping, verification, execution, and rollback autonomously across entire projects. It runs locally and consists of a server for orchestration and a CLI for workflow requests.
• Diffblue Cover CoreAutonomously writes and maintains human-like JUnit or TestNG unit tests that are comprehensive, compile, run, and accurately validate code behavior for entire applications.
• Diffblue Cover OptimizeSpeeds up the time required to run Java unit tests by running only the tests in a project that are relevant to a code change, minimizing local and CI test execution time.
• Diffblue Cover ReportsProvides reporting and visualization insights for teams into the state of unit testing in terms of coverage levels, coverage risk, testability, and actionable insights to improve code quality.
• Diffblue Cover CLIProcesses entire codebases autonomously, including legacy Java 8 and 11 codebases. It describes the current behavior of the code and writes an entire test suite that runs, compiles, and is guaranteed to be correct.
• Diffblue Cover PipelineAnalyzes pull/merge requests and fills any gaps, updating tests to maintain maximum test coverage.
• Diffblue Cover IDE pluginAssists developers in writing and updating their unit tests as they modernize code.

Key capabilities

• Automated unit test generation
• Automated unit test maintenance and management
• Integration with existing AI coding platforms (GitHub Copilot, Claude Code)
• Autonomous operation across entire codebases
• Support for Java (8, 11, 17, 21, 25) and Python (3.9+)
• Reinforcement learning technology for guaranteed correct code
• On-premise deployment for security and privacy
• Outcome-based pricing (per net new line of coverage)
• Verification framework ensuring tests compile and pass
• Smart regression test execution
• Continuous testing capabilities
• Code documentation through unit tests
• Regression detection and impact measurement
• Code coverage visualization and quality reporting
• Suggests and automates code fixes for testability

Use cases

• Automating the generation, maintenance and management of Java unit tests
• Legacy code & application modernization
• Code coverage improvement
• Full/hybrid cloud migration
• Regression testing
• Continuous testing
• DevOps and test automation
• Accelerating unit testing
• Refactoring code
• Avoiding regressions
• Understanding application behavior and complex legacy code
• Improving developer productivity and velocity
• Meeting code coverage targets
• Streamlining PR workflows

AI approach

Diffblue uses reinforcement learning to generate, maintain, and manage Java and Python unit tests. It acts as an AI agent that orchestrates existing AI coding platforms (like GitHub Copilot and Claude Code) to achieve comprehensive test coverage, verify output, and prepare pull requests. The technology is designed to be autonomous, deterministic, and accurate, operating locally or on-premise to ensure security and data privacy.

Tech named: AI, ML, Reinforcement Learning, AI agent, LLM, code completion tools

Industries served

• Software Development
• Financial Services
• Healthcare
• Aerospace and Defense

What it says sets it apart

• 10x faster than GitHub Copilot for unit testing
• 250x faster than manual test writing
• 26x productivity boost compared to a developer using Copilot/other coding assistants
• Orchestrates AI coding tools for comprehensive coverage
• Guarantees tests run, compile, and are correct using reinforcement learning
• Operates autonomously across entire codebases
• On-premise deployment for enhanced security and data privacy
• Outcome-based pricing based on net new lines of coverage
• Supports legacy Java 8 and 11 codebases
• Built-in verification framework for generated tests
• Does not require constant developer monitoring or rework
• Automated fixing and clean-up of errors
• Patented reinforcement learning technology with high precision and low error rate
• Continuously maintains and updates tests when changes are made
• Does not train on customer code, ensuring IP safety