Wraps test runners in your project and returns structured JSON instead of terminal output. Auto-detects pytest, jest, vitest, and mocha, so you don't configure which framework you're using. The agent gets typed test results (pass/fail counts, individual test status, error messages) without parsing ANSI codes or fragile text formats. Part of the larger Pare toolkit that does the same for git, docker, npm, and other CLI tools. Reach for this when you want an agent to run tests and act on the results programmatically, like filing issues for failures or generating coverage reports. Token savings hit 80% on verbose test output compared to raw terminal text.
claude mcp add --transport stdio io.github.dave-london-pare-test -- npx -y @paretools/test