Quick Start
Get from zero to your first findings in under 60 seconds. This guide covers the fastest path.
The zero-config way
The absolute fastest start. Navigate to any directory and run s0 with no arguments:
cd your-project/
s0System 0 will create an instance, choose a starting analysis angle, and run one pass. Findings appear immediately in your terminal.
The explicit way
For more control, create a named instance with a specific goal:
s0 init my-audit \
--goal 'Find security vulnerabilities and code quality issues' \
--target /path/to/your/project \
--mode observeThen run analysis passes:
s0 run my-audit --cycles 5 --verboseThe --verbose flag shows each finding as it is discovered, with severity scores.
Review what it found
# See findings, gaps, and dimensions
s0 observe my-audit
# See the analysis angles it's developed
s0 questions my-audit
# Check spending
s0 budget my-auditGraduate to fixing
Once you trust the analysis, switch from observing to fixing:
# Plans only, you review
s0 mode my-audit --set suggest
# Queue fixes for your approval
s0 mode my-audit --set supervised
s0 run my-audit --cycles 3
s0 approve my-audit --all
# Fully autonomous
s0 mode my-audit --set auto
s0 run my-audit --cycles 10Set it and forget it
Install as a background daemon for continuous analysis:
# Run every 10 minutes
s0 install my-audit --interval 600
# Or run continuously (back-to-back)
s0 install my-audit --continuousThe daemon runs via launchd (macOS) or systemd (Linux). It survives reboots, respects budget limits, and logs everything.
What happens during a pass?
- Select: The system picks an analysis angle based on what has been most productive so far
- Analyze: The LLM examines your target against that angle
- Collect: Findings (issues with severity scores) and gaps (blind spots) are recorded
- Grow: New analysis angles emerge from discovered gaps
- Act: If the mode allows, fixes are planned, refined, and executed
- Log: Everything goes to the append-only ledger
Key commands to remember
| Command | What it does |
|---|---|
s0 | Analyze current directory (auto-creates instance) |
s0 init NAME | Create a named instance |
s0 run NAME | Run analysis passes |
s0 observe NAME | See what it found |
s0 status | List all instances |
s0 install NAME | Run as background daemon |
Next steps
- Full command reference for every option
- Key concepts to understand how the system thinks
- Configuration to tune budgets, thresholds, and backends
- Real-world examples for specific use cases