First Steps¶
Run your first AI workflow and see results in under 5 minutes.
Your First Workflow¶
Let's run a security audit on some code. This workflow scans for vulnerabilities and provides actionable recommendations.
Option 1: CLI (Fastest)¶
# Scan your source directory
attune workflow run security-audit --path ./src
# Or scan current directory
attune workflow run security-audit --path .
Example output:
Security Audit Results
======================
Status: completed
Findings: 3
[HIGH] SQL query vulnerable to injection
File: src/database.py:45
Fix: Use parameterized queries
[MEDIUM] Hardcoded API key in config
File: src/config.py:12
Fix: Move to environment variable
[LOW] Missing input validation
File: src/handlers/user.py:28
Fix: Validate user input before processing
Cost: $0.0850
Option 2: Python¶
from attune.workflows import SecurityAuditWorkflow
import asyncio
async def audit():
workflow = SecurityAuditWorkflow()
result = await workflow.execute(path="./src")
print(f"Success: {result.success}")
print(result.summary or result.final_output)
print(f"\nCost: ${result.cost_report.total_cost:.4f}")
asyncio.run(audit())
Try More Workflows¶
Attune AI includes 17 built-in workflows. Here are the most popular:
| Workflow | Command | What It Does |
|---|---|---|
| Security Audit | attune workflow run security-audit |
Find vulnerabilities |
| Bug Prediction | attune workflow run bug-predict |
Predict likely bugs |
| Code Review | attune workflow run code-review |
Tiered code analysis |
| Test Generation | attune workflow run test-gen |
Generate missing tests |
| Release Prep | attune workflow run release-prep |
Pre-release checklist |
| Dependency Check | attune workflow run dependency-check |
Find outdated deps |
| Perf Audit | attune workflow run perf-audit |
Find bottlenecks |
# List all available workflows
attune workflow list
# Get help for a specific workflow
attune workflow run security-audit --help
Understanding the Output¶
Every workflow returns:
| Field | Description |
|---|---|
success |
True if the workflow completed without error |
summary |
Human-readable summary of the run |
final_output |
The workflow's primary result (analysis text) |
cost_report |
API costs and cache hit rate (.total_cost, ...) |
metadata |
Timing, model used, etc. |
Cost Tracking¶
All telemetry data stays local in ~/.attune/telemetry/.
What Just Happened?¶
When you ran the security audit:
- File scanning - The workflow read your source files
- Tiered analysis - Simple files used a cheap model, complex ones used a capable model
- Pattern matching - Known vulnerability patterns were detected
- Report generation - Results were formatted and returned
This is the tiered model approach - automatically routing to the right model based on task complexity.
Interactive Wizards (Claude Code)¶
If you're using Claude Code, try the guided wizards for a more interactive experience:
/wizard run debug # Step-by-step debugging
/wizard run security # Guided security audit
/wizard run refactor # Refactoring with analysis
/wizard run test-gen # Test generation wizard
/wizard run release-prep # Release readiness check
Wizards ask questions, adapt to your answers, and produce structured action plans. They use the same workflows under the hood but guide you through the process interactively.
Next Step¶
Now that you've run a workflow, it's time to Choose Your Path based on how you want to use the framework.
See Also¶
- Choose Your Path - Decide between Quick Start, Tutorial, or Deep Dive
- Installation Guide - Package options and configuration
- MCP Integration - Connect to Claude Code for IDE integration
- Multi-Agent Coordination - Build agent teams
- Configuration Reference - Customize framework behavior