# Quick Start: Think-MCP Testing
Get from zero to evaluated in 5 minutes.
---
## Prerequisites
- think-mcp server installed (`npm install think-mcp`)
- Claude Code with MCP configured
---
## Step 1: Verify MCP Connection
Open Claude Code and check:
```
Show me all available MCP tools
```
You should see:
- `mcp__think-mcp__trace`
- `mcp__think-mcp__model`
- `mcp__think-mcp__pattern`
- ... (11 total)
---
## Step 2: Quick Smoke Test
Test one tool:
```
Use mcp__think-mcp__trace with:
{
"thought": "Testing the MCP connection",
"thoughtNumber": 1,
"totalThoughts": 1,
"nextThoughtNeeded": false
}
```
If this works, you're ready!
---
## Step 3: Run Full Evaluation
### Option A: Using Tester Agent
```bash
claude code --agent think-mcp-tester
```
This runs:
- 47 schema validation tests
- LLM-as-Judge semantic evaluation
- 33 tool accuracy scenarios
### Option B: Manual Testing
For individual tool testing:
```javascript
// trace tool
mcp__think-mcp__trace({
thought: "Step-by-step analysis",
thoughtNumber: 1,
totalThoughts: 3,
nextThoughtNeeded: true
})
// model tool
mcp__think-mcp__model({
modelName: "first_principles",
problem: "Why does our build take 15 minutes?"
})
// pattern tool
mcp__think-mcp__pattern({
patternName: "api_integration",
context: "Integrating multiple payment providers"
})
```
---
## Step 4: Generate Report
```bash
claude code --agent think-mcp-reporter
```
Creates `think-mcp-eval-report.md` with:
- Executive summary
- Schema quality analysis
- Semantic quality scores
- Tool accuracy metrics
- Recommendations
---
## Expected Results
| Metric | Target | Actual |
|--------|--------|--------|
| Schema Quality | >= 95% | 100% |
| Semantic Score | >= 0.85 | 0.89 |
| Tool Accuracy F1 | >= 0.90 | 0.97 |
| Avg Response | < 500ms | 199ms |
---
## Troubleshooting
### "Tool not found"
1. Check think-mcp server is running
2. Verify Claude Code MCP config
3. Restart Claude Code
### "Type validation error"
Ensure inputs match expected schema:
```javascript
// Wrong
{ thought: 123 }
// Correct
{ thought: "Test", thoughtNumber: 1, totalThoughts: 1, nextThoughtNeeded: false }
```
---
## Files Reference
```
test-results/
├── think-mcp-eval-results.jsonl # Test results
├── think-mcp-eval-results-semantic.jsonl # Semantic scores
├── tool-accuracy-scenarios.jsonl # Accuracy test cases
├── tool-accuracy-results.jsonl # Accuracy results
└── think-mcp-eval-report.md # Comprehensive report
```
---
## Summary Checklist
- [ ] MCP server verified
- [ ] Smoke test passed
- [ ] Full evaluation run
- [ ] Report generated
- [ ] Results reviewed
