Skip to Content

start

The start command is used to initiate a new test job.

Usage

bytesalt start <command>

Arguments

  • <command> (required): A brief, descriptive command for what you want to run. This prompt is used to give the AI agent a high-level overview of the test objective.

Note: The bytesalt CLI automatically looks for a .bytesalt/context directory in your current working directory. Any files in this directory will be uploaded and provided to the agent as context.

Be as specific as possible with your prompt. Very broad tasks can take a significant amount of time and credits. Instead of “Test my app”, try “Test the user registration flow and verify the welcome email is mentioned in the UI.”

Options

  • --tunnel: Use this flag to enable the secure tunnel for testing localhost or internal networks. See tunnel to read more.

  • --context [name=][@]source: Provide additional dynamic context (can be repeated).

    • source can be a local file path, a literal string, or - for stdin.
    • Use @ to specify that the source is a file path or stream.
    • name is the optional filename used inside the context archive.

    Examples:

    • --context @file.txt (AI sees ‘file.txt’)
    • --context pr-diff.patch=@<(git diff main) (Process substitution)
    • --context "note.txt=This is a literal string" (Literal text)
  • --output type=<full|summary>,format=<pdf|markdown>,dest=<path>: Specify output artifacts to download.

    • Can be repeated for multiple outputs.
    • type: Report content type (full or summary).
    • format: File format (pdf, markdown, or json).
    • dest: Local destination path (relative or absolute).
  • --fail-if "expression": Define a custom failure condition using a CEL expression. If the expression evaluates to true, the command exits with code 1.

    • Default (if not specified): result.issues.critical > 0 || result.issues.high > 0 (fails on critical or high severity issues)
    • Example: result.issues.total > 0 (fail on any issues)
    • Example: result.issues.critical > 0 (fail only on critical issues)
    • See Policy Expressions for the result schema and more examples.

Example

bytesalt start "Test registration" --tunnel --fail-if "result.issues.critical > 0"

This command will start the job and, once finished, fail if any critical issues are found. The custom --fail-if overrides the default behavior.

Last updated on