Structured, token-efficient test runner output for AI agents. Up to 80% fewer tokens than raw test CLI output.
Part of the Pare suite of MCP servers. Auto-detects pytest, jest, vitest, and mocha.
| Tool | Description |
|---|---|
run |
Run tests, returns pass/fail counts and failure details |
coverage |
Run tests with coverage, returns per-file summary |
playwright |
Run Playwright tests with structured pass/fail output |
npx -y @paretools/testAdd to your MCP client config:
{
"mcpServers": {
"pare-test": {
"command": "npx",
"args": ["-y", "@paretools/test"]
}
}
}run output:
{
"framework": "vitest",
"passed": 46,
"failed": 1,
"skipped": 0,
"total": 47,
"failures": [
{
"name": "parseOutput > handles empty input",
"file": "src/__tests__/parsers.test.ts",
"message": "expected true to be false"
}
]
}- Tool process timeout (
run,coverage,playwright):300_000ms(5 minutes) - Integration test MCP call timeout:
300_000ms - Vitest file/test timeout for this package:
300_000ms
This package runs real downstream suites (including @paretools/git integration tests), so lower global limits like 120s can fail due to cumulative runtime on slower CI/Windows runners.
pnpm test:unitruns fast parser/formatter/helper testspnpm test:integrationruns MCP integration + tool-params testspnpm test:fidelityruns the long fidelity suite
The default pnpm test runs these batches sequentially for earlier failure and clearer triage.
Works with any MCP-compatible client: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code / GitHub Copilot, Cline, Roo Code, Zed, Continue.dev, Gemini CLI, OpenAI Codex
