Skip to content

Choose Your Path

You've installed the framework and run your first workflow. Now choose the approach that fits your needs.

Four Ways to Use Attune AI

Path Best For Complexity
CLI Power User Quick tasks, automation, CI/CD Simple
MCP Integration Claude Desktop, conversational workflow building Simple
Workflow Developer Custom automations, Python integration Moderate
Meta-Orchestration Complex tasks, multi-agent teams Advanced

Path 1: CLI Power User

Best for: Quick tasks, shell scripts, CI/CD pipelines

Use the empathy CLI to run pre-built workflows without writing Python.

Key Commands

# Run workflows
attune workflow run security-audit --path ./src
attune workflow run bug-predict --path ./src
attune workflow run release-prep --path .

# Track costs
attune telemetry show
attune telemetry savings --days 30

Next Steps

Path 2: MCP Integration

Best for: Claude Desktop users, conversational workflow building

Connect to Claude Desktop or any MCP-compatible client for guided workflow creation.

Quick Setup

Add to Claude Desktop config:

{
    "mcpServers": {
        "socratic": {
            "command": "python",
            "args": ["-m", "attune.socratic.mcp_server"],
            "env": {"ANTHROPIC_API_KEY": "your-key"}
        }
    }
}

Then ask Claude to help you build workflows conversationally.

Next Steps

Path 3: Workflow Developer

Best for: Custom automations, integrating AI into Python apps

Use the Python API to run and build workflows.

Using Built-in Workflows

from attune.workflows import SecurityAuditWorkflow
import asyncio

async def audit():
    workflow = SecurityAuditWorkflow()
    result = await workflow.execute(target_path="./src")
    print(f"Found {len(result.findings)} issues")

asyncio.run(audit())

Next Steps

Path 4: Meta-Orchestration

Best for: Complex tasks needing multiple AI agents

Describe what you want and let the framework compose agent teams.

from attune.orchestration import MetaOrchestrator

orchestrator = MetaOrchestrator()
plan = orchestrator.analyze_and_compose(
    task="Review code for security and suggest performance improvements",
    context={"path": "./src"}
)
result = await orchestrator.execute(plan)

Next Steps

Still Not Sure?

If you want to... Start with...
Run quick tasks from terminal CLI
Use Claude Desktop MCP Integration
Build custom Python apps Workflow Developer
Orchestrate complex multi-agent tasks Meta-Orchestration

Most users start with CLI or MCP. Move to Workflow Developer when you need custom logic, and Meta-Orchestration when tasks get complex.