FluxLoop for Claude Code
Test AI agents directly from your IDE. Just talk naturally—FluxLoop handles the rest.
⭐ The Core: Agent Test Skill
Forget commands. Just ask Claude.
"test my agent"
"generate test data"
"run a simulation"
"set up fluxloop"
The Agent Test Skill automatically:
- ✅ Checks your setup state
- ✅ Guides you through installation
- ✅ Generates test data and personas
- ✅ Runs tests against your agent
- ✅ Uploads results to the cloud
- ✅ Shows you what to fix
No commands to memorize. No manual steps.
Quick Start
1. Install the Plugin
/plugin install Fluxloop-AI/fluxloop-claude-plugin
2. Just Talk
You: "I want to test my customer support agent"
That's it. The skill takes over and guides you through everything.
How the Skill Works
Context-First Workflow
You: "test my agent"
↓
[Skill checks your state]
↓
Claude: "Here's your current setup:
- CLI: ✓ Installed
- Auth: ✓ Logged in
- Project: customer-support
- Bundle: v3 (50 inputs)
Would you like to run tests now?"
↓
You: "yes"
↓
[Skill runs tests, shows results]
You're always in control. Nothing runs without your approval.
Auto-Activation Keywords
The skill activates when you mention:
| What You Say | What Happens |
|---|---|
| "test my agent" | Run tests |
| "generate test data" | Create synthetic inputs |
| "run simulation" | Execute agent simulation |
| "set up fluxloop" | Full installation guide |
| "evaluate responses" | Analyze agent quality |
Example Conversation
You: "test my chatbot with difficult customers"
Claude: I'll help you test your agent.
📊 Current State:
- Project: customer-support
- Scenario: support-bot
- Available bundles:
1. v3 (stress-test, 50 inputs)
2. v2 (basic, 10 inputs)
Which bundle would you like to use, or should I generate new test data
with difficult customer personas?
You: "generate new with difficult customers"
Claude: Creating personas focused on difficult customers...
[Executes: fluxloop personas suggest]
✓ Generated 5 personas:
- angry_customer
- confused_user
- demanding_professional
- impatient_caller
- skeptical_buyer
Now synthesizing test inputs...
[Executes: fluxloop inputs synthesize]
✓ 30 test inputs created!
✓ Published as bundle v4
Running tests...
[Executes: fluxloop sync pull + fluxloop test]
📊 Results: 25/30 passed (83%)
⚠️ 5 failures detected
Would you like me to analyze the failures?
Why Skill > Commands?
| Approach | Experience |
|---|---|
| Skill | "test my agent" → Everything automatic |
| Commands | /fluxloop:pull → /fluxloop:test → manual steps |
The skill is:
- Smarter: Knows your context, adapts to your state
- Safer: Always asks before executing
- Simpler: No syntax to remember
Slash Commands (Manual Alternative)
For when you need direct control:
| Command | Description |
|---|---|
/fluxloop:setup | First-time setup |
/fluxloop:test | Run tests |
/fluxloop:smoke | Quick smoke test |
/fluxloop:pull | Pull test data |
/fluxloop:status | Check status |
💡 Recommendation: Use natural language instead. The skill handles edge cases automatically.
Requirements
- FluxLoop Account: alpha.app.fluxloop.ai
- Python 3.8+
💡 The skill installs everything for you! Just say "set up fluxloop" and it handles CLI installation, login, and project setup automatically.
Architecture
┌─────────────────────────────────────┐
│ You: "test my agent" │
└────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ Agent Test Skill │
│ (Context-aware, guides workflow) │
└────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ FluxLoop CLI │
│ (Executes commands) │
└────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ FluxLoop Web │
│ (Cloud storage & analysis) │
└─────────────────────────────────────┘
Best Practices
Let the Skill Guide You
Good: "I want to test edge cases"
Less helpful: "/fluxloop:synthesis" (manual)
Be Specific About Goals
Good: "test my order cancellation flow with angry customers"
Less helpful: "test it"
Ask for Analysis
"Why did these tests fail?"
"What patterns do you see?"
"How can I improve my agent?"
Troubleshooting
Skill Not Activating?
Include activation keywords:
✗ "check this"
✓ "test this agent"
Need Manual Control?
Use slash commands:
/fluxloop:status
/fluxloop:test --smoke
What's Next?
- Agent Test Skill ⭐ - Full skill documentation
- Commands Reference - Manual commands
- Installation - Setup details
- FluxLoop Web - Cloud dashboard
Need help? Just ask: "help me with fluxloop" — the skill will guide you!