MCP Server for Developers

Run your tests from
inside your AI assistant

RunCrucible is an MCP server that lets Claude, Cursor, and Windsurf detect and run your test suite — so your AI can see failures and fix them directly.

Start for $19/month View on GitHub →
$ npm install -g runcrucible-mcp

What it does

Six tools. Zero configuration.

Add RunCrucible to your MCP config and your AI assistant gets a complete testing workflow — detect, run, fix, cover, generate, and find flaky tests.

🔍

detect_test_framework

Scans your repo and identifies Jest, Vitest, pytest, or Mocha automatically from your config files.

run_tests

Executes your test suite and returns full output — stdout, stderr, exit code — so your AI can read failures and fix them without leaving the editor.

🎯

run_tests_with_context

Runs tests and on failure automatically reads the failing test files and their source counterparts — everything needed to fix in one call.

📊

get_coverage

Runs coverage and returns a per-file report highlighting everything below your threshold. Your AI knows exactly what to write tests for.

✍️

generate_test_file

Reads a source file and any existing tests nearby. Returns full context so your AI writes an accurate, complete test suite in one shot.

🎲

detect_flaky_tests

Runs your suite multiple times and flags tests with inconsistent results — the ones that waste hours of debugging because nobody built the detector.


Setup

Three lines to configure.

Add this to your MCP config and restart your client. That's it.

claude_desktop_config.json / .cursor/mcp.json
// Claude Code, Cursor, Windsurf — same config format
{
  "mcpServers": {
    "runcrucible": {
      "command": "runcrucible-mcp",
      "env": {
        "RUNCRUCIBLE_LICENSE_KEY": "your-license-key"
      }
    }
  }
}

Supported

Works with your stack.

Framework is auto-detected. No flags, no extra config.

Jest Vitest Mocha pytest More coming

Pricing

Simple, flat pricing.

Pro

$19

per month · cancel anytime

Get Started