Skip to main content
Every command accepts the flags below on top of its own. The command reference lists only what is unique to each command, so these apply everywhere.

Global flags

Environment variables

Agent mode

Agent mode is for anything that is not a person: a script, a CI job, an AI agent. It prints JSON to stdout, drops colour and chrome, and gives you a stable contract to parse.
On success, the raw data payload goes to stdout with exit code 0. Pipe it straight into jq:
On failure, the same envelope the API uses goes to stdout, and the exit code is non-zero:
code and message are always present. status appears when the failure came from the API rather than from the CLI itself. hint appears when there is something specific to suggest. A CLI-side failure (a bad flag, an unreachable API) carries the code cli_error.
A production write still asks for confirmation in agent mode. A non-interactive shell cannot answer, so the write fails closed. Pass --yes on any write you expect to run unattended.

Safety

The CLI is read-first. Every write passes three guards.

Sending a request body

Write commands accept individual fields as flags, a whole JSON body, or a body file. You can combine them: field flags merge on top of --body.

Raw requests

If you need an endpoint the CLI does not wrap as a named command yet, call it directly. api handles auth, the base URL, and the envelope, and it respects the same write guards.
--query repeats, one key=value per flag. Bodies come from --body or --body-file, as above.

Command reference

Every command’s arguments and flags in full.

Authentication

Where the key and config live, and how CI reads them.

Troubleshooting

When a write is blocked and you cannot see why.

CLI overview

Install the tool, and see what it covers.