Use with Claude Code

A ready-to-paste prompt that teaches Claude Code (or any AI agent) to drive BizOS through the CLI.

Paste the prompt block below into Claude Code, ChatGPT, or any terminal-capable agent. It tells the agent which bizos commands exist, that credit-spending commands need --yes, and to use --json for parsing. This is the headline use case for BizOS.

Why the CLI is AI-first

The bizos CLI is a small, predictable command surface where every command supports --json. That makes it trivial for an AI agent to call a command, parse the result, and decide the next step. Billing is enforced server-side, so an agent can operate autonomously without ever spending a credit by accident — paid actions are gated behind an explicit --yes.

Prerequisites

The CLI is installed and you are logged in — see Quickstart.
The agent runs in an environment where bizos is on the PATH. For unattended runs, export BIZOS_API_KEY so no interactive login is needed.

Copy-paste prompt for Claude Code

Paste this verbatim. It is self-contained.

You can operate my BizOS account through the `bizos` CLI, which is installed and
authenticated on this machine. BizOS is a thin client over a billing-enforced API,
scoped to one organization.

Rules:
- ALWAYS pass `--json` so you can parse output reliably.
- Listing and reading are FREE. Two commands SPEND 1 credit each and REQUIRE `--yes`:
  `bizos tasks run <id> --yes` and `bizos creative generate "<prompt>" --yes`.
  Never spend a credit unless I explicitly asked for that action.
- If a command returns HTTP 402 / {"error":"out_of_credits"}, STOP and tell me to
  top up at app.bizos.lol/billing. Do not retry.
- If a command returns 401, tell me to run `bizos login`.

Command surface (run `bizos <group> --help` for flags):
- Identity:   bizos whoami
- Tasks:      bizos tasks list | bizos tasks create "<text>" | bizos tasks run <id> --yes  (⚠️ 1 credit)
- Docs:       bizos docs list [--folder <id>] [--all] | bizos docs get <id>
              | bizos docs create --title "<t>" [--content "<md>"|--file <path>] [--kind <k>] [--folder <id>] [--summary "<text>"]
- Leads:      bizos leads list
- Creative:   bizos creative generate "<prompt>" [--size 1024x1024] [--quality <q>]
              [--angle <a>] [--format <f>] [--label "<l>"] --yes                (⚠️ 1 credit)
- Landing:    bizos landing list | bizos landing create [--business-name "<s>"] [--subdomain <s>] [--page-type <s>] [--spec <json>] [--strategy <json>] [--no-render] [--publish]
- Meta (RO):  bizos meta campaigns | bizos meta insights [--range <range>]
- Billing:    bizos wallet | bizos billing
- Notifs:     bizos notifications | bizos notifications read <id>

Workflow: to run a task, first `bizos tasks list --json` to find the id, confirm with
me, then `bizos tasks run <id> --yes`. Before any credit-spending command, check
`bizos wallet --json` and tell me the balance.

A shorter prompt (one-liner)

For quick sessions:

Drive my BizOS account with the `bizos` CLI (installed + logged in). Always use --json.
Reading is free; `bizos tasks run` and `bizos creative generate` cost 1 credit and need
--yes — never spend a credit unless I ask. On 402, stop and tell me to top up.

Example agent session

A typical autonomous loop an agent would run:

# 1. Confirm identity and credit balance (free)
bizos whoami --json

# 2. Find a task to run (free)
bizos tasks list --json

# 3. Run it — explicit, paid, gated by --yes (⚠️ 1 credit)
bizos tasks run tsk_123 --yes --json

# 4. Read the result document (free)
bizos docs list --json

Safety notes for agents

Credits are real money. Treat --yes as a spend authorization. The CLI also refuses to spend without it.
Failures refund. If a paid action fails server-side, the credit is automatically refunded — the agent does not need to reconcile.
One org per key. Everything the agent does is scoped to the org the key belongs to; it cannot reach another org.

Continue to the Commands reference for full flag lists and example JSON for every command.