Diffblue’s Testing Agent autonomously generates unit tests across a codebase, working on individual classes and methods or in batch mode across entire repositories without file-by-file prompting, and extends existing AI platforms such as GitHub Copilot CLI and Claude Code rather than replacing them. Every generated test is run through a built-in verification framework before it counts: it must compile, pass, and add measurable coverage, so a team never has to manually review a batch of generated tests for basic correctness. Diffblue Ltd., spun out of the University of Oxford in 2016 by founders Daniel Kroening and Peter Schrammel, counts Citi, MongoDB, BNY and SanDisk among its customers and supports Java 8 through 25 and Python 3.9 and later.
Diffblue writes and verifies tests that compile and pass, but a passing, coverage-adding test is not automatically a well-designed one, so a team should still review generated tests for whether they check meaningful behavior rather than trivial code paths, especially before relying on them as the sole regression safety net for critical logic. The billing model is based on net new lines of coverage measured against a captured baseline, so a team should understand how that baseline is set before assuming a specific dollar figure for a given project.
Pricing starts at $1,500 for 5,000 net new lines of coverage added to a codebase, working out to $0.30 per line, a figure a customer can independently verify with JaCoCo, Cobertura or any standard coverage tool. Larger and multi-repository deployments are available as custom enterprise packages with volume pricing, SSO/SAML, dedicated support, SLA guarantees and on-premises deployment for regulated industries. A free trial is available to evaluate results before committing to a paid package.
