Technical Guide

AgencyAI CLI

Manage your consulting products, orders, and analytics from the terminal. Same backend as MCP and the dashboard.

Installation#

The CLI is included with the AgencyAI platform. Clone the repo and run it with npx tsx:

git clone https://github.com/your-org/agent-operator-os.git
cd agent-operator-os
npm install

# Run the CLI
npm run cli -- --help

Or run directly with npx tsx:

npx tsx src/cli/index.ts --help

Authentication#

The CLI uses the same API keys as MCP and the platform API. Get your key from Dashboard → Settings → Integrations.

# Log in with your API key
agencyai auth login --key op_live_your_key_here

# Check who you're logged in as
agencyai auth status

# Log out
agencyai auth logout

You can also set the OPERATOR_API_KEY environment variable instead of logging in:

export OPERATOR_API_KEY=op_live_your_key_here

Quick Start#

1. List Your Products

agencyai products list

Shows a table with title, type, price, and status for all your products.

2. Create a Product

agencyai products create

Interactive prompt for title, description, price, type, and at least 3 intake questions. Creates in draft status.

3. View Orders

# List recent orders
agencyai orders list

# Get full details for an order
agencyai orders get <order-id>

# Fulfill an order
agencyai orders fulfill <order-id> --content '{"result": "Done"}'

Command Reference#

`agencyai auth`

Command	Description
`auth login --key <KEY>`	Validate and save API key
`auth status`	Show logged-in expert info
`auth logout`	Clear saved credentials

`agencyai products`

Command	Flags	Description
`products list`	--status, --limit	Table of your products
`products get <id>`	—	Full product details
`products create`	—	Interactive product creation
`products publish <id>`	—	Set status to published
`products retire <id>`	—	Set status to retired

`agencyai orders`

Command	Flags	Description
`orders list`	--status, --limit	Recent orders table
`orders get <id>`	—	Full order + intake responses
`orders fulfill <id>`	--content, --notes	Mark as delivered

Other Commands

Command	Description
`storefront view`	Show storefront URL and published product count
`storefront open`	Open storefront in browser
`analytics overview`	Orders by status, revenue, top products
`keys list`	List API keys with prefix, name, last used
`keys create --name <name>`	Generate a new API key (shown once)
`keys revoke <id>`	Revoke an API key

MCP Integration#

The CLI and MCP server share the same backend — they use the same API keys, query the same Supabase database, and execute the same business logic. Think of them as two interfaces to the same system:

CLI

Best for scripting, batch operations, CI/CD workflows, and quick terminal-based management. Interactive mode for product creation.

MCP Server

Best for AI agent integration. Connect from Claude, Cursor, or any MCP-compatible tool to let AI manage your products and orders autonomously.

# MCP server uses the same auth
agencyai auth login --key op_live_xxx   # CLI
# MCP also validates op_live_ keys via the same function

Dashboard#

The third interface is the web dashboard at /dashboard. All three interfaces — CLI, MCP, and Dashboard — talk to the same Supabase database and enforce the same ownership rules.

CLI

Terminal

MCP

AI Agents

Dashboard

Web UI

Configuration#

CLI config is stored at ~/.agencyai/config.json:

{
  "apiKey": "op_live_...",
  "serverUrl": "http://localhost:3010"
}

The CLI connects directly to Supabase using Drizzle ORM — no web server required. Set DATABASE_URL in your .env file.