Skip to main content
Usage:
Options:
  • --help: Show this message and exit.
Commands:
  • configure
  • create
  • start
  • validate: Validate the output of an LLM against a…
  • watch

​
configure

Usage:
Options:
  • --enable-metrics / --disable-metrics: Opt out of anonymous metrics collection. [default: enable-metrics]
  • --token TEXT: API Key for Guardrails. If not provided, you will be prompted for it.
  • --enable-remote-inferencing / --disable-remote-inferencing: Opt in to remote inferencing. Deprecated — Guardrails-hosted remote inference is being discontinued on August 6, 2026. [default: enable-remote-inferencing]
  • --clear-token: Clear the existing token from the configuration file.
  • --help: Show this message and exit.

​
create

Usage:
Options:
  • --validators TEXT: A comma-separated list of validators. [required]
  • --name TEXT: The name of the guard to define in the file.
  • --filepath TEXT: The path to which the configuration file should be saved. [default: config.py]
  • --dry-run / --no-dry-run: Print out the validators to be installed without making any changes. [default: no-dry-run]
  • --help: Show this message and exit.
Installing validators. Validators are published on public PyPI. Install them with pip, e.g. pip install guardrails-ai-regex-match, and uninstall with pip uninstall guardrails-ai-regex-match.

​
start

Usage:
Options:
  • --env TEXT: An env file to load environment variables from.
  • --config TEXT: A config file to load Guards from.
  • --port INTEGER: The port to run the server on. [default: 8000]
  • --help: Show this message and exit.

​
validate

Validate the output of an LLM against a rail spec. Usage:
Arguments:
  • RAIL: Path to the rail spec. [required]
  • LLM_OUTPUT: String of llm output. [required]
Options:
  • --out TEXT: Path to the compiled output directory. [default: .rail_output]
  • --help: Show this message and exit.

​
watch

Usage:
Options:
  • --plain / --no-plain: Do not use any rich formatting, instead printing each entry on a line. [default: no-plain]
  • --num-lines INTEGER: Print the last n most recent lines. If omitted, will print all history. [default: 0]
  • --follow / --no-follow: Continuously read the last output commands [default: follow]
  • --log-path-override TEXT: Specify a path to the log output file.
  • --help: Show this message and exit.