Official command-line interface for Autousers — run UX
evaluations, kick off autouser runs, and inspect calibration from your
terminal. Same ak_live_* API key as @autousers/mcp.
npm install -g @autousers/cliOr invoke without installing:
npx -y @autousers/cli --helpRequires Node.js 20 or newer.
@autousers/cli runs in two modes, auto-detected from how you invoke it:
| Command | Mode |
|---|---|
autousers |
TUI (interactive Ink-based UI, requires a TTY) |
autousers <subcommand> (e.g. eval list) |
Plain CLI |
autousers --no-tui |
Plain CLI (force) |
autousers in CI / non-TTY |
Plain CLI (auto-detected) |
autousers --help / autousers --version |
Plain CLI |
The TUI is the friendly default for interactive use. Plain mode is for
scripting, CI, and one-shot commands. Every TUI capability has a
plain-mode equivalent — see autousers --help.
The auto-detection rule lives in shouldUseTUI(): TUI iff stdout is a
TTY and CI is unset and --no-tui isn't passed and there's
no positional subcommand on argv. A bare autousers lands in the TUI;
anything else (including --help / --version) falls through to
commander.
Interactive — launch the TUI:
autousersScripted — list running evaluations as JSON:
autousers eval list --json | jq '.[] | select(.status == "Running")'CI — non-interactive eval run with cost preview:
# Step 1: dry-run to see what it would cost — no tokens spent.
autousers eval run <id>
# Step 2: confirm and queue the autouser runs.
autousers eval run <id> --commitForce plain mode in an interactive shell:
autousers --no-tui # prints help, no TUI
autousers eval list --no-tuiautousers loginOpens your default browser, signs you in to autousers.ai, and persists a
short-lived OAuth access token + refresh token to ~/.autousers/config.json.
Same UX as claude /mcp — no API key to copy. The CLI refreshes the token
silently when it expires.
If you're on SSH or a no-display environment:
autousers login --no-browserThe CLI prints the auth URL; open it on a machine that has a browser, sign in, and the localhost callback resolves once you complete consent.
Mint a key at app.autousers.ai/settings/api-keys — keys are shown once, so paste it into a secrets manager immediately.
autousers login --key ak_live_...Or supply via env (best for CI):
export AUTOUSERS_API_KEY=ak_live_...
autousers eval list # no `login` step neededThe CLI walks these in order, first match wins:
--key ak_live_...— explicit per-invocation flagAUTOUSERS_API_KEY=ak_live_...— environment variable- OAuth access token in
~/.autousers/config.json(auto-refreshed) apiKeyin~/.autousers/config.json
To override the API host (self-hosted / local dev):
export AUTOUSERS_BASE_URL=http://localhost:3000Default: https://app.autousers.ai.
autousers login # browser OAuth flow
autousers whoami # confirm the active user / team
autousers eval list # list your evaluations
autousers eval get <id> # detailed eval view
autousers autouser list # list calibrated personas
autousers template list # list rating templates
autousers usage --range 30d # quota + token spend rollup
autousers logout # revoke + clear local credentialsEvery command supports --json for stable machine-readable output:
autousers eval list --json | jq '.[] | select(.status == "Running")'Run any subcommand with --help for its specific flags.
| Variable | Default | Description |
|---|---|---|
AUTOUSERS_API_KEY |
— | Bearer token (ak_live_...) |
AUTOUSERS_BASE_URL |
https://app.autousers.ai |
Override the API host |
NO_COLOR |
— | 1 disables ANSI color |
Global flags: --json, --quiet, --no-color, --key, --base-url.
Exit codes: 0 OK · 1 generic error · 2 auth · 3 not found · 4
validation · 64 usage error.
- autousers.ai/help/cli — full CLI guide
@autousers/mcp— Autousers inside Claude / Cursor / ChatGPT via the Model Context Protocol- Mint an API key
- License — MIT