Flinchify

MCP allowing your Agent to request human validation on webapps and apps.

GitHub

flinchify-mcp

Human usability testing for AI agents via Model Context Protocol.

Real humans test your app. Your AI reads the results and fixes the bugs.

Works with Claude Desktop, Cursor, OpenAI (ChatGPT), Gemini, Grok, and any MCP-compatible client.

Quick Start

npx flinchify-mcp

Get your API key at flinchify.com/dashboard → API Keys tab.

Setup

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "flinchify": {
      "command": "npx",
      "args": ["flinchify-mcp"],
      "env": {
        "FLINCHIFY_API_KEY": "fk_your_key_here"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "flinchify": {
      "command": "npx",
      "args": ["flinchify-mcp"],
      "env": {
        "FLINCHIFY_API_KEY": "fk_your_key_here"
      }
    }
  }
}

OpenAI (ChatGPT)

Add to your MCP tool configuration:

{
  "command": "npx",
  "args": ["flinchify-mcp"],
  "env": {
    "FLINCHIFY_API_KEY": "fk_your_key_here"
  }
}

Gemini

{
  "mcpServers": {
    "flinchify": {
      "command": "npx",
      "args": ["flinchify-mcp"],
      "env": {
        "FLINCHIFY_API_KEY": "fk_your_key_here"
      }
    }
  }
}

Grok

{
  "mcpServers": {
    "flinchify": {
      "command": "npx",
      "args": ["flinchify-mcp"],
      "env": {
        "FLINCHIFY_API_KEY": "fk_your_key_here"
      }
    }
  }
}

Windsurf / Any MCP Client

Any client that supports the Model Context Protocol can use Flinchify. Point it at npx flinchify-mcp with your API key as an environment variable.

Available Tools

create_test

Create a human usability test job. Real humans will test the given URL, follow your specified flow, and report back with issues, screen recordings, and friction notes.

Example prompt: "Have 5 real humans test my app at https://myapp.com — they should sign up, create a project, and try to invite a teammate. Budget $10 per tester."

Parameters:

  • url (required) — URL to test
  • flow — What testers should do
  • testers — Number of testers (default: 3, max: 100)
  • budget_per_tester — Per-tester budget in USD (default: $10, min: $5)
  • tasks — Array of specific tasks (alternative to flow)
  • target_audience — Ideal tester profile
  • time_limit_hours — Hours to complete (default: 24)

get_test_results

Get severity-ranked issues, tester feedback, screen recording URLs, and completion status.

Example prompt: "Check the results from test ft_127"

list_tests

List all your test jobs with status and results summary.

check_credits

Check your credit balance. Credits let you create tests instantly without checkout interruptions.

How It Works

  1. Sign up at flinchify.com and get an API key
  2. Buy credits ($50–$500 packs) or pay per test via Stripe checkout
  3. Your AI agent creates test jobs — real humans are matched and test your app
  4. Results come back structured: severity-ranked issues, feedback, recordings
  5. Your AI reads the results and can fix bugs automatically

Pricing

  • Minimum $5/tester — you set the price
  • Credit packs: Starter ($50), Growth ($150), Scale ($500)
  • Credits = instant test creation, no checkout interruption
  • No credits? Falls back to Stripe checkout link

Links

License

MIT

Verwandte Server