# Workflow MCP Testing Guide
## Prerequisites
Before testing, ensure:
1. NCP is built: `npm run build`
2. Anthropic API key is configured
## Step-by-Step Testing
### Step 1: Verify Workflow MCP is Loaded
```bash
# Check if workflow MCP is discovered
ncp find "workflow"
# Should show:
# workflow:execute
# workflow:execute-custom
# workflow:execute-task
# workflow:generate-workflow
# workflow:list-workflows
```
If not found, check:
```bash
# Verify file exists
ls -la src/internal-mcps/examples/workflow.mcp.ts
# Rebuild
npm run build
```
### Step 2: Configure Anthropic API Key
**Option A: Add to NCP profile** (recommended):
Edit `~/.ncp/all.json`:
```json
{
"mcpServers": {
"workflow": {
"type": "internal",
"env": {
"ANTHROPIC_API_KEY": "_USE_SECURE_STORAGE_"
}
}
}
}
```
Then run any workflow tool - you'll be prompted to enter the key.
**Option B: Set environment variable**:
```bash
export ANTHROPIC_API_KEY="sk-ant-..."
```
### Step 3: Test Simple Task Execution
**Test 1: Basic intelligent task**
```bash
ncp run workflow:execute-task \
task="Write a haiku about programming"
```
Expected output:
```json
{
"task": "Write a haiku about programming",
"result": "Code flows like water...",
"usage": { "input_tokens": 25, "output_tokens": 50 },
"model": "claude-3-5-sonnet-20241022"
}
```
**Test 2: Task with context**
```bash
ncp run workflow:execute-task \
task="Summarize these GitHub issues" \
context='{"issues":["Bug in login","Feature request: dark mode","Performance issue"]}'
```
### Step 4: Test Workflow Listing
```bash
ncp run workflow:list-workflows
```
Expected output:
```json
{
"workflows": [
{
"name": "daily-standup",
"description": "Daily standup: Check issues, summarize, post to Slack",
"requiredContext": ["slackWebhook", "repo"],
"recommendedSchedule": "0 9 * * 1-5"
},
// ... more workflows
]
}
```
### Step 5: Test Workflow Generation
```bash
ncp run workflow:generate-workflow \
description="Every hour, check Hacker News front page, find AI-related posts, summarize them, and save to a file"
```
Expected: Claude generates a workflow definition with steps.
### Step 6: Test Predefined Workflow (Simulation)
Since predefined workflows need actual MCPs (GitHub, Slack, etc.), let's test the structure:
```bash
# This will execute but may fail at tool calling (expected)
ncp run workflow:execute \
workflowName="daily-standup" \
context='{"slackWebhook":"https://test.webhook","repo":"test/repo"}'
```
**Expected behavior**:
- ✅ Workflow loads correctly
- ✅ Steps are identified
- ⚠️ Tool calls may fail (if GitHub/Slack MCPs not configured)
- ✅ State tracking works
### Step 7: Test Custom Workflow
Create a simple test workflow:
```bash
ncp run workflow:execute-custom \
workflow='{
"name": "test-workflow",
"description": "Simple test workflow",
"steps": [
{
"name": "Step 1: Think",
"description": "Generate a random fact about cats"
},
{
"name": "Step 2: Analyze",
"description": "Explain why cats are interesting"
}
]
}'
```
**Expected output**:
```json
{
"workflowName": "test-workflow",
"status": "completed",
"results": [
{
"step": "Step 1: Think",
"success": true,
"output": "Cats can rotate their ears 180 degrees...",
"timestamp": "2024-01-15T10:00:00Z"
},
{
"step": "Step 2: Analyze",
"success": true,
"output": "Cats are interesting because...",
"timestamp": "2024-01-15T10:00:05Z"
}
],
"startTime": "2024-01-15T10:00:00Z",
"endTime": "2024-01-15T10:00:05Z"
}
```
## Integration Testing with Other MCPs
### Test with AI MCP
```bash
ncp run workflow:execute-custom \
workflow='{
"name": "ai-test",
"description": "Test AI integration",
"steps": [
{
"name": "Generate Text",
"description": "Use AI to write a short poem",
"tool": "ai:complete",
"parameters": {
"prompt": "Write a short poem about technology",
"provider": "anthropic"
}
}
]
}'
```
**Note**: This requires AI MCP to be configured with API keys.
### Test with Database MCP
```bash
ncp run workflow:execute-custom \
workflow='{
"name": "db-test",
"description": "Test database workflow",
"steps": [
{
"name": "Create Test DB",
"description": "Create a test database table",
"tool": "database:create-table",
"parameters": {
"dbPath": "./test-workflow.db",
"tableName": "workflow_tests",
"columns": {"id": "INTEGER PRIMARY KEY", "name": "TEXT", "timestamp": "TEXT"}
}
},
{
"name": "Insert Test Data",
"description": "Insert a test record",
"tool": "database:insert",
"parameters": {
"dbPath": "./test-workflow.db",
"tableName": "workflow_tests",
"data": {"name": "Test Workflow", "timestamp": "2024-01-15"}
}
}
]
}'
```
## Testing with Schedule MCP Integration
### Step 1: Test Workflow Manually
```bash
# Create a simple workflow that just logs
ncp run workflow:execute-task \
task="Return the current date and time with a friendly message"
```
### Step 2: Schedule It
```bash
# Schedule to run every 5 minutes for testing
ncp run schedule:add \
name="test-workflow" \
command="ncp run workflow:execute-task task='Log test execution at current time'" \
schedule="*/5 * * * *" \
description="Test workflow execution every 5 minutes"
```
### Step 3: Verify Schedule
```bash
# List schedules
ncp run schedule:list
# Should show your test-workflow
```
### Step 4: Monitor Execution
Wait 5 minutes and check if it ran:
```bash
# Check schedule status/logs (if supported)
ncp run schedule:get name="test-workflow"
```
### Step 5: Remove Test Schedule
```bash
ncp run schedule:remove name="test-workflow"
```
## Troubleshooting Tests
### Issue 1: "workflow not found"
**Symptom**:
```
Error: Tool not found: workflow:execute
```
**Solutions**:
```bash
# 1. Rebuild
npm run build
# 2. Check file exists
ls src/internal-mcps/examples/workflow.mcp.ts
# 3. Check for TypeScript errors
npm run build 2>&1 | grep workflow
# 4. Verify SimpleMCP loader is working
ncp find "ai" # Test with another example MCP
```
### Issue 2: "ANTHROPIC_API_KEY not configured"
**Symptom**:
```
Error: ANTHROPIC_API_KEY not configured.
```
**Solution**:
```bash
# Option 1: Set environment variable
export ANTHROPIC_API_KEY="sk-ant-..."
# Option 2: Add to profile
# Edit ~/.ncp/all.json and add workflow MCP config
# Verify
echo $ANTHROPIC_API_KEY
```
### Issue 3: Workflow executes but tool calls fail
**Symptom**:
```json
{
"step": "Call GitHub",
"success": false,
"error": "Tool github:list-issues not found"
}
```
**Explanation**: This is expected if referenced MCPs aren't configured.
**Solutions**:
1. Configure the referenced MCP (e.g., GitHub)
2. Or test with a workflow that doesn't call external tools
3. Or modify workflow to use available MCPs only
### Issue 4: Conditional steps not working
**Test conditional logic**:
```bash
ncp run workflow:execute-custom \
workflow='{
"name": "conditional-test",
"steps": [
{
"name": "Step 1",
"description": "Always runs"
},
{
"name": "Step 2",
"description": "Should skip",
"condition": "false"
},
{
"name": "Step 3",
"description": "Should run",
"condition": "true"
}
]
}' \
context='{}'
```
Expected: Step 2 skipped, Steps 1 and 3 executed.
## Performance Testing
### Test 1: Single Step Latency
```bash
time ncp run workflow:execute-task \
task="Say hello"
```
Expected: ~2-5 seconds (depends on Claude API latency)
### Test 2: Multi-Step Workflow Latency
```bash
time ncp run workflow:execute-custom \
workflow='{
"name": "multi-step",
"steps": [
{"name": "Step 1", "description": "First step"},
{"name": "Step 2", "description": "Second step"},
{"name": "Step 3", "description": "Third step"}
]
}'
```
Expected: ~6-15 seconds (3 steps × 2-5 seconds each)
### Test 3: Token Usage
Check how many tokens workflows consume:
```bash
ncp run workflow:execute-task \
task="Generate a comprehensive report about AI trends" | jq '.usage'
```
## Test Checklist
Before considering Workflow MCP ready:
- [ ] `ncp find "workflow"` shows all 5 tools
- [ ] `execute-task` works with simple task
- [ ] `list-workflows` returns predefined workflows
- [ ] `generate-workflow` creates valid workflow from description
- [ ] `execute-custom` runs multi-step workflow
- [ ] Error handling works (test with invalid API key)
- [ ] Integration with Schedule MCP works
- [ ] State tracking shows correct status
- [ ] Conditional steps execute correctly
- [ ] Context variables are accessible in steps
## Quick Test Script
Save this as `test-workflow.sh`:
```bash
#!/bin/bash
echo "Testing Workflow MCP..."
echo "1. Testing discovery..."
ncp find "workflow" | grep -q "workflow:execute" && echo "✓ Discovery works" || echo "✗ Discovery failed"
echo "2. Testing simple task..."
ncp run workflow:execute-task task="Say hello in French" > /dev/null 2>&1 && echo "✓ Simple task works" || echo "✗ Simple task failed"
echo "3. Testing workflow listing..."
ncp run workflow:list-workflows | grep -q "daily-standup" && echo "✓ Workflow listing works" || echo "✗ Workflow listing failed"
echo "4. Testing custom workflow..."
ncp run workflow:execute-custom workflow='{"name":"test","steps":[{"name":"test","description":"test step"}]}' > /dev/null 2>&1 && echo "✓ Custom workflow works" || echo "✗ Custom workflow failed"
echo "Done!"
```
Run it:
```bash
chmod +x test-workflow.sh
./test-workflow.sh
```
## Next Steps
After basic testing works:
1. **Test real workflows** with actual MCP integrations
2. **Schedule workflows** with Schedule MCP
3. **Monitor execution** over time
4. **Optimize** based on token usage and latency
5. **Create custom workflows** for your specific use cases
Happy testing! 🚀
