Output & automation

ayx is built to be driven by scripts and agents, not just typed at a prompt. Two things make that work: a uniform JSON envelope and a dry-run-by-default safety gate.

JSON output

Add -o json for machine-readable output. It’s a global flag, so it can appear before or after the subcommand:

ayx -o json one workflows list

Every JSON response uses the same envelope:

{
  "ok": true,
  "command": "one.workflows.list",
  "message": "…",
  "timestamp_utc": "…",
  "data": {}
}

Branch on ok, read data for the result. Failures use the same shape with ok: false, add an error_code, and are written to stderr instead of stdout. command names the leaf that ran on both, so a log line can be traced back to the invocation that produced it.

Dry-run by default

Mutating commands print what they would do and exit 0 unless you pass --apply. A pipeline can safely run the dry-run form against production:

# Preview — no changes; capture the planned request
ayx -o json one workflows delete <id>

# Commit, non-interactively
ayx one workflows delete <id> --apply --yes

--yes skips the confirmation prompt that destructive commands show in a terminal — required for CI and pipes.

Exit codes

• 0 — success, or a dry-run that completed
• non-zero — the command failed; read message (and data) in the envelope for why

Agent-friendly by design

Because every command speaks the same JSON envelope and gates destructive actions behind an explicit flag, ayx makes a clean tool surface for AI agents: predictable output to parse, and a safety rail so an agent can’t change remote state without an explicit --apply.