agent-cards companies wizard --help
Usage: agent-cards companies wizard [options]
Run an AI agent that implements Agentcard (OAuth + MCP) into the repo at this
path
Options:
--path <dir> Target repo directory (defaults to the current directory)
-y, --yes Skip the consent prompt (required in agent mode: explicit
consent to modify files)
--agent Non-interactive agent mode: no prompts, plain streaming
output, machine-readable result line (auto-enabled under
CLAUDECODE/CI or without a TTY)
--email <email> Email for sign-in when no session exists (interactive:
prefills the sign-in prompt)
--code <code> Agent mode: sign-in code from the email (only for
backends on code sign-in, after a code_required exit;
interactive runs prompt for the code)
--org <id-or-name> Company to provision under (agent mode: required when the
account has several; interactive: skips the picker)
--app-url <url> Your app's base URL for the OAuth callback (default
http://localhost:3000)
--app-name <name> Name for the company (when created) and OAuth client
(default: the repo directory name)
--verbose Include tool inputs and outputs in the stream
--engineer Route Agentcard uncertainty to the LIVE forward-deployed
engineer instead of guessing from the static playbook
(needs `engineer init` + a running `engineer connect` for
this repo)
-h, --help display help for command
Agent mode (for AI agents / CI):
agent-cards companies wizard --agent --yes [--email you@co.com] [--org <id>] [--app-url http://localhost:3000]
(also reachable as: npx agent-cards-admin wizard --agent --yes — the old name forwards here)
- auto-enabled without --agent when CLAUDECODE or CI is set, or a TTY is missing
on either side; AGENT_CARDS_WIZARD_MODE=agent|interactive overrides auto-detection
- never prompts; anything it needs and doesn't have exits 2 — the result line's
"hint" names the flag to re-run with and what to ask your user when the value
isn't derivable (the emailed sign-in code, which company, the app's base URL)
- sign-in: prints "magic link sent to <email>" and polls up to 15 minutes (env
AGENT_CARDS_WIZARD_LOGIN_TIMEOUT_MS overrides); re-runs RESUME the pending
sign-in instead of re-emailing. Backends on code sign-in exit 2 with
code_required — finish with --email <email> --code <code>
- runs take 5-20 minutes end to end: use a generous timeout or run in background
- streams plain progress lines (add --verbose for tool detail); if a connect link
needs a user tap it appears in the stream — open it to let the run continue
- the LAST line is machine-readable: AGENTCARD_WIZARD_RESULT {"ok":true,...}
- exit codes: 0 success · 1 integration/unexpected error · 2 missing input/consent
error codes: consent_required, login_required, code_required, invalid_email,
invalid_path, invalid_app_url, invalid_app_name, org_ambiguous, org_not_found, root_unsupported,
integration_agent_error, unexpected_error
- --engineer additionally requires a live engagement; missing pieces exit 2 with
engineer_engagement_required, engineer_engagement_not_active, or
engineer_daemon_not_running (the hint names the command a human must run)