Reference
CLI Reference
Every flag the llm-scanner command understands.
| Flag | Required | Default | Description |
|---|---|---|---|
| --target | Yes | - | URL or Ollama model name to test |
| --target-type | Yes | - | url or ollama |
| --judge-model | Yes | - | Local Ollama model used as the AI evaluator |
| --categories | No | LLM01-LLM09 | Comma-separated OWASP categories to test. Example: LLM01,LLM07 |
| --severity | No | all | Minimum severity to include: critical high medium low info |
| --api-key | No | None | Bearer token for authenticated endpoints. Never logged or printed. |
| --output-dir | No | ./reports | Directory for saved Markdown, JSON, and HTML reports |
| --format | No | terminal only | Comma-separated output formats: md, json, html |
| --include-dos-tests | No | off | Include LLM10 (Unbounded Consumption) probes. Only use against targets you own. |
| --log-level | No | INFO | Logging verbosity: DEBUG INFO WARNING ERROR |
| --log-file | No | None | Optional path to write structured JSON log lines |
See Output Formats for what each --format value produces, and OWASP LLM Top 10 for what --categories can filter on.