crucible

danhannah94/crucible
★ 0 stars JavaScript 🤖 AI/LLM Updated today
Agentic QA framework for AI coding agents — visual verification, baseline management, and a project-agnostic MCP server. Gives your agents eyes.
View on GitHub → Try with Claude — $10 free →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "crucible": {
      "command": "npx",
      "args": [
        "-y",
        "danhannah94/crucible"
      ]
    }
  }
}

README Excerpt

> **Agentic QA framework that gives AI coding agents eyes.** A project-agnostic MCP server for visual verification, baseline management, and repeatable browser testing — built for agents that use your app's own MCP tools (or generic browser tools as a fallback) to interact, and Crucible to *see*. The market is full of "AI helps you write Playwright scripts faster." Crucible is the opposite: **the agent is the test runner.** No scripted steps, no brittle selectors. The agent uses domain-specific

Tools (14)

CRUCIBLE_ADAPTERCRUCIBLE_BASELINE_ROOTCRUCIBLE_SMOKE_PROJECTCRUCIBLE_SMOKE_SPECCRUCIBLE_SMOKE_URLCRUCIBLE_TEST_ENV_URLapprove_baselineclickcompare_screenshotslist_baselinesnavigateread_window_globalrun_scriptscreenshot_page