Diffblue Cover, now offered as the Diffblue Testing Agent, generates, verifies and fixes unit tests autonomously across enterprise codebases. Rather than prompting one test at a time, it runs coverage analysis, repairs build problems, plans tests, generates them in parallel, verifies the output and prepares a pull request. Every delivered test is checked to compile and pass. Pricing is outcome-based: packages start at $1,500 for 5,000 net new lines of coverage, about $0.30 per line, and larger enterprise packages carry volume pricing.
The audience is enterprise engineering and QA teams that need high regression coverage on legacy or fast-changing code, especially Java 8, 11, 17, 21 and 25 plus Python 3.9 and later. Platform groups modernizing old applications use it to build a safety net before refactoring, and organizations that already approve an AI coding assistant can run Diffblue on top of it without evaluating a new model vendor. Regulated industries can ask about on-premises deployment, SSO and SLA options.
In an internal benchmark across eight Java repositories totalling 31,069 coverable lines, all starting at 0% coverage, Diffblue reports 80.7% average line coverage against 32.3% for a senior developer working with a general AI coding agent. Teams point the CLI at a repository in batch mode, or write tests for individual classes in the IDE through task mode. Typical jobs include raising coverage before a migration, verifying AI-generated code does not break existing behavior, and measuring results with JaCoCo or Cobertura. Support for additional coding agents is listed as coming soon.
