CodiumAI Cover Agent is a tool that aims to help increase code coverage efficiently by automatically generating qualified tests that enhance a project’s existing test suites. It is presented alongside PR-Agent, another CodiumAI project, in a survey of coding agents applied to software engineering.
Roadmap
What is known about it comes from its published roadmap of planned features, which records current implementation status alongside each item. The central item is automatically generating unit tests for software projects with AI models, aiming at comprehensive test coverage and quality assurance — an item the roadmap annotates “similar to Meta”. Under that item it lists generating tests for different programming languages, handling a wide variety of testing scenarios, producing a behaviour analysis of the code under test and generating tests from it, and checking tests for flakiness, for example by running them five times as suggested by TestGen-LLM.
A second group widens the test-generation pains it covers: generating new tests focused on the changeset of a pull request, and running over an entire repository to try to enhance all its existing test suites. A third group is about usability: connectors for GitHub Actions, Jenkins, CircleCI, Travis CI and others; integration with databases, APIs, OpenTelemetry and other data sources to extract relevant inputs and outputs for test generation; and a settings file.