Axolotl

Axolotl analyzes a pull request's code changes, figures out what should be tested, and then actually tests it in a real browser against the running application, instead of relying only on eyeballing a diff and an existing test suite. A QA session can start from a PR number, uncommitted changes, specific files, or a chat prompt.

Its pipeline uses tree-sitter AST parsing and ripgrep pattern search to understand the code structure before writing tests, searches the web for testing practices specific to the project's tech stack to generate targeted test cases, and can run the project's own lint, typecheck, and test scripts on the changed files first, forwarding any real CI failures as context into the test planner so the generated plan is not built around hallucinated bugs. After a reviewer approves the generated test plan, Axolotl injects temporary log markers, runs real browser-based end-to-end tests while monitoring both frontend UI and backend output, and produces an evidence-backed pass/fail verdict per test case. It requires an Anthropic API key, and web search during test-case generation uses a separate search API key.