🎯 SlopWatch - AI Accountability MCP Server

Stop AI from lying about what it implemented! Track what AI claims vs what it actually does.

🚀 What's New in v2.6.0

✨ Ultra-Minimal Responses - 90% less verbose output
🔄 Combined Tool - Single call instead of 2 separate tools
⚡ Seamless Workflow - Perfect for AI pair programming

🤔 Why SlopWatch?

Ever had AI say "I've added error handling to your function" but it actually didn't? Or claim it "implemented user authentication" when it just added a comment?

SlopWatch catches AI lies in real-time.

⚡ Quick Start

Option 1: Smithery (Recommended - 1 click install)

1. Visit smithery.ai/server/@JoodasCode/slopwatch
2. Click "Install to Cursor" or "Install to Claude"
3. Done! ✨

Option 2: NPM Install

npm install -g slopwatch-mcp-server

🔧 Configuration

Cursor IDE

Add to your MCP settings:

{
  "mcpServers": {
    "slopwatch": {
      "command": "npx",
      "args": ["slopwatch-mcp-server"]
    }
  }
}

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "slopwatch": {
      "command": "npx",
      "args": ["slopwatch-mcp-server"]
    }
  }
}

🎮 How to Use

Method 1: Combined Tool (Recommended ⭐)

Perfect for when AI implements something and you want to verify it:

// AI implements code, then verifies in ONE call:
slopwatch_claim_and_verify({
  claim: "Add input validation to calculateSum function",
  originalFileContents: {
    "utils/math.js": "function calculateSum(a, b) { return a + b; }"
  },
  updatedFileContents: {
    "utils/math.js": "function calculateSum(a, b) {\n  if (typeof a !== 'number' || typeof b !== 'number') {\n    throw new Error('Invalid input');\n  }\n  return a + b;\n}"
  }
});

// Response: "✅ PASSED (87%)"

Method 2: Traditional 2-Step Process

For when you want to claim before implementing:

// Step 1: Register claim
slopwatch_claim({
  claim: "Add error handling to user login",
  fileContents: {
    "auth.js": "function login(user) { return authenticate(user); }"
  }
});
// Response: "Claim ID: abc123"

// Step 2: Verify after implementation
slopwatch_verify({
  claimId: "abc123",
  updatedFileContents: {
    "auth.js": "function login(user) {\n  try {\n    return authenticate(user);\n  } catch (error) {\n    throw new Error('Login failed');\n  }\n}"
  }
});
// Response: "✅ PASSED (92%)"

🛠️ Available Tools

| Tool | Description | Response | |------|-------------|----------| | slopwatch_claim_and_verify | ⭐ Recommended - Claim and verify in one call | ✅ PASSED (87%) | | slopwatch_claim | Register what you're about to implement | Claim ID: abc123 | | slopwatch_verify | Verify implementation matches claim | ✅ PASSED (92%) | | slopwatch_status | Get your accountability stats | Accuracy: 95% (19/20) | | slopwatch_setup_rules | Generate .cursorrules for automatic enforcement | Minimal rules content |

💡 Real-World Examples

Example 1: API Endpoint Enhancement

// AI says: "I'll add rate limiting to your API endpoint"

slopwatch_claim_and_verify({
  claim: "Add rate limiting to user registration endpoint",
  originalFileContents: {
    "routes/auth.js": "app.post('/register', async (req, res) => {\n  const user = await createUser(req.body);\n  res.json(user);\n});"
  },
  updatedFileContents: {
    "routes/auth.js": "const rateLimit = require('express-rate-limit');\n\nconst registerLimit = rateLimit({\n  windowMs: 15 * 60 * 1000, // 15 minutes\n  max: 5 // limit each IP to 5 requests per windowMs\n});\n\napp.post('/register', registerLimit, async (req, res) => {\n  const user = await createUser(req.body);\n  res.json(user);\n});"
  }
});

// Result: ✅ PASSED (94%) - AI actually implemented rate limiting!

Example 2: React Component Update

// AI says: "I'll add loading states to your component"

slopwatch_claim_and_verify({
  claim: "Add loading spinner to UserProfile component",
  originalFileContents: {
    "components/UserProfile.jsx": "export function UserProfile({ userId }) {\n  const user = fetchUser(userId);\n  return <div>{user.name}</div>;\n}"
  },
  updatedFileContents: {
    "components/UserProfile.jsx": "export function UserProfile({ userId }) {\n  const [user, setUser] = useState(null);\n  const [loading, setLoading] = useState(true);\n\n  useEffect(() => {\n    fetchUser(userId).then(userData => {\n      setUser(userData);\n      setLoading(false);\n    });\n  }, [userId]);\n\n  if (loading) return <div className='spinner'>Loading...</div>;\n  return <div>{user.name}</div>;\n}"
  }
});

// Result: ✅ PASSED (89%) - AI added proper loading state!

Example 3: Catching AI Lies

// AI claims: "I've added comprehensive error handling"
// But actually just added a comment...

slopwatch_claim_and_verify({
  claim: "Add comprehensive error handling to payment processing",
  originalFileContents: {
    "payment.js": "function processPayment(amount) {\n  return stripe.charges.create({ amount });\n}"
  },
  updatedFileContents: {
    "payment.js": "function processPayment(amount) {\n  // TODO: Add error handling\n  return stripe.charges.create({ amount });\n}"
  }
});

// Result: ❌ FAILED (15%) - Busted! AI only added a comment, no actual error handling

📊 Check Your AI's Accuracy

slopwatch_status();
// Response: "Accuracy: 87% (26/30)"

Track how often your AI actually implements what it claims vs just talking about it!

🎯 Automatic Enforcement

Generate .cursorrules to automatically enforce SlopWatch usage:

slopwatch_setup_rules({ project_path: "/path/to/project" });

This creates rules that require AI to use SlopWatch for every implementation claim.

🔥 Benefits

• ✅ Catch AI lies before they make it to production
• ✅ Build trust in AI pair programming
• ✅ Improve code quality through verification
• ✅ Track accuracy over time
• ✅ Ultra-minimal responses don't clutter your chat
• ✅ Works with any MCP-compatible IDE

🌟 Why Developers Love SlopWatch

"Finally caught my AI claiming it added tests when it just added a comment!"
— @developer_mike

"The combined tool is a game-changer. One call instead of two!"
— @sarah_codes

"87% accuracy rate revealed my AI was lying way more than I thought."
— @tech_lead_jane

🚀 Getting Started

1. Install: Use Smithery (1-click) or NPM
2. Configure: Add to your IDE's MCP settings
3. Use: Start with slopwatch_claim_and_verify for best experience
4. Monitor: Check your accuracy with slopwatch_status

🔗 Links

• 🏠 Homepage: smithery.ai/server/@JoodasCode/slopwatch
• 📦 NPM Package: slopwatch-mcp-server
• 🐛 Issues: GitHub Issues
• 💬 Creator: @mindonthechain

📄 License

MIT License - Free for everyone! 🎉

Made with ❤️ by @mindonthechain
Stop AI slop. Start AI accountability. 🎯